CEU eTD Collection In partial fulfilment of the requirements for the degree of thedegree for oftherequirements In partialfulfilment Comparing Causes of Party System Institutionalisation Electoral Volatility in Old and New Democracies: New and in Old Volatility Electoral Department of Political Science of Political Department Central European University European Central University European Central Supervisor: Zsolt Enyedi Supervisor: Benjamin Rowe Jones Rowe Benjamin Budapest, Hungary Budapest, Masters ofMasters Arts Submitted to (2012) By CEU eTD Collection factors. elite-driven and institutional both by hampered been has stability system party rule, military yearsleftby of legacy alongside the historical twenty that party Brazilian system concluding increase in electoral volatility.Finally, this paper also provides an in depth case study of the presidential serveregimes toundermine party system institutionalisation causing a significant original finding of the paper is that unlike parliamentary or presidential regimes, semi- most the Additionally, more regimes recently. inaugurated those volatility than electoral in inaugurated democracies experiencingearlier periods lowerlevel asignificantly of find that in period the took withplacedemocratisation which isthedecisivefactor, those system institutionalisation, age of democracy does not play a determining role. Instead, we this study finds thatcontrary findingsto the much of of traditionalthe literature on party through both statistical and case study analyses. Based on the most extensive data assembled, paper attempts fillto gapin this examining thecauses of party system institutionalisation scholars have considered what factors may or may not contribute to this phenomenon. This Although interest in party system institutionalisation remains high within the discipline, few Abstract i CEU eTD Collection this thesis would nothave been possible. contributions these Without comments. insightful and patience his for Littvay Levente thank Eleanor Neff Powell and Joshua Taylorin sharing their data with me. Finally, Iwould like to throughout this Iam project. also grateful for generosity the Mainwaring,shown by Scott I wouldlike to mythank supervisor ZsoltEnyedi for his support and wise guidance Acknowledgements ii CEU eTD Collection 4.1 Ranges ...... 4.1 28 FINDINGS 27 AND ...... RESULTS 4 3.6 Independent3.6 VariablesHypothesesand ...... 17 Dataset 3.5 ...... 16 Measurement3.4 ...... 15 DependentVariable3.3 ...... 13 Operationalisation3.2 ...... 12 Mixed-Methods...... 3.1 11 HYPOTHESES ...... 11 AND DATA, DESIGN, 2.5 Causes ofParty System Institutionalism –TheStatethe Art of ...... 7 Party2.4 System Institutionalisation ...... 6 Whatis Institutionalisation?2.3 ...... 6 theyMatter? do Why 2.2 ...... 5 WhatPartyis a 2.1 System? ...... 4 ...... 4 SYSTEMS 2 PARTY 1 INTRODUCTION ...... 1 ...... VII ABBREVIATIONS OF LIST FIGURES ...... OF VI LIST TABLES ...... OF V LIST CONTENTS III ...... OF TABLE II ...... ACKNOWLEDGEMENTS ...... I ABSTRACT 3.6.5 Volatility3.6.5 Educationand ...... 25 3.6.4Volatility Regimeand Type ...... 23 Volatility3.6.3Fragmentationand ...... 21 3.6.2Volatility theand Economy ...... 20 Volatility3.6.1Time and ...... 17 Table of Contents iii CEU eTD Collection 6 CONCLUSIONS 68 CONCLUSIONS ...... 6 ...... 59 STUDY 5 CASE Discussion 4.9 ...... 49 Multi-level4.8 Regression Analysis ...... 47 Multivariate4.7 Regression Analysis ...... 41 BivariateCorrelation4.6 Analysis ...... 38 Fragmentation4.6 ...... 37 Education 4.5 ...... 37 RegimeType 4.4 ...... 36 Volatility4.3 ...... inOrder 34 Trendsover4.2 time ...... 30 iv CEU eTD Collection Table.11 Table.10 Table.9 Table.8 Table.7 Table.6 Table.5 Table.4 Table.3 Table.2 Table.1 List of Tables –Determinants randomintercepts ofelectoralvolatility model –Determinants bysingle election ofelectoralvolatility volatility electoral average of –Determinants – Lowestaverage volatility bycountry –Highestaverage by volatility country 4 byelectionnumber group volatility – Average –Average by volatility election number by –Averagegroup volatility – Volatility ranges – Volatility by group Position of the Parties Positionofthe in ExecutiveandLegislativeOffice(%) elections between volatility electoral of Correlations - v CEU eTD Collection Figure.9 Figure.8 Figure.7 Figure.6 Figure.5 Figure.4 Figure.3 Figure.2 Figure.1 List of Figures – GDP per capita and electoral volatility scatter –GDPpercapitaandelectoralvolatility plot life –Average expectancy school (years) Group4b –Averagebyelectionnumber volatility Group4 –Averagebyelectionnumber volatility 1,2,and3 –Average forGroups volatilitybyelection number –Average volatilitybyelection number - Party shares Rio de Janeiro delegation to Chamber of Deputies 1982 – Chamber ofDeputies19822002 Janeiro to - PartysharesRiode delegation 1 distanceformodel - Cooks – Average ENP vi CEU eTD Collection PTB - PSD – da ReconstruçãoNacional PRN –Partido NIMD –Netherlands Institute for MultipartyDemocracy MDB – Renovadora Nacional ARENA -Aliança of Abbreviations List Partido Trabalhista Brasileiro Trabalhista Partido Partido da Social Democracia Brasileira Democracia Social da Partido Movimento Democrático Brasileiro Democrático Movimento vii CEU eTD Collection several respects. It ties together a number of heretofore unconnected strands of analysis and in of strands analysis and heretofore unconnected Itties anumber of several together respects. system institutionalisation,also whilst providinga unique contribution to literaturethe in This paperre-evaluates the findingsof handfulthe of into causes previousthe studies of party research. of area fruitful a provide well may institutionalisation system successful party to that can contribute factors ofthe analysis if robust moreas a democracy; and across countries with similar levels of economic development, it would seem institutionalisation across geographical countries across regions; with similar experiences of 2010p.1;Bértoa Mainwaring& 2007 p.156). Zoco Given wide the indivergences levels of causesp.155), the of party haveinstitutionalisation system often (Casal been overlooked and debate,sparked vibrant particularly last heated inthe 2007 & decade(Mainwaring Zoco times has at that science political of area isan institutionalisation system party Although new struggling democracies. or Consequentlyp.317). study the of party system institutionalisation is particularly salient in modern democracies (Mainwaring & Scully p.1;Mair2005 1997p.199; Kitschelt2001 of anessential constituting requirement and stability, thereby accountability legitimacy, butalso onlynot operate ensures a well-institutionaliseddemocracies helps system party that consolidation Ágh(Sartori 1990 p.317; 1998 p.206). Many leading argue theorists democratic and institutionalised party system, is therefore, an important and validmeasure of democratic andmaintain and accountability Mairdemocratic (Bartolini 2001p.328).A well functioning political landscape (Mainwaring & 2005 p.3); Scully ensure theorganisation of government, inFurthermore, Hale,2006,p.1). parties, and by extension systems, helpparty shape the ‘unimaginable’ ‘unthinkable’ ‘unworkable’(Aldrich) (Schattschneider)(Weber), and (Cited is political parties without discuss democracy to that extent tothe demands collective into Political parties play a crucial in role individualaggregating citizens’ interests and passions 1 Introduction 1 CEU eTD Collection particularly low levels of party system institutionalisation. system party of levels low particularly influence types levels regime presidential or parliamentary neither that show regression multi-level of electoral and The of in results both much multivariate the of literature. the hypothesised typically manner in the not but country, a by experienced volatility electoral of level volatility.the in determining Arguably mostthe original finding of this study is that regime type can play importantan role Instead,party system institutionalisation. semi-presidentialmore recently, even when controlling anumber for of variables hypothesised toinfluence inaugurated regimes those than volatility electoral of level lower a significantly experience 1990s).Thisthe democracies in findingsuggests those that that inaugurated earlier periods regimes in transition democratic experienced that countries (those analysis the within democratisation are‘period effect’ thesis 2007 (Mainwaringp.166) by &Zoco including afourth wave of showndemocracies can be more widely inferred. Secondly, study this further substantiates the to producenew in institutionalisation system party of levels in explaining besignificant to find mythat arenot findings due to sample bias andindicatesfurther thatthe causes selection I emphasisunderwent that in transition on 1990s. This those suggests the wider rangeof cases includingtwenty-one cases, amuch more diverse range withnewof democracies, a particular party system institutionalisation in fourty-seven countries; this study extends this figure by Sequences and theStabilizationof Interparty Competition, levels Previously of Mainwaringeconomic development). and2007 article, Zoco’s been previously considered region, of (in terms number, with experience democracy, and system institutionalisation. causesof for analysesof the party subsequentcomparative new benchmarkestablished a The analysis also has It drawn. be can conclusions which upon basis isrobust more a appliedprovides so doing to a much wider range of cases than has 2 was uniquein analysing causesof Political CEU eTD Collection question drawn conclusions the by literature. previous the into call therefore and robust are here findings my that believe to reason good have I study, studies. sizeand included However,given the several important diversity data myinthe of change in levels of party system institutionalisation. This previousthe contradicts findings of Finally, study is this novel infinding that neither level education of nor GDPproduce any 3 CEU eTD Collection systems ineffective In theirhighly governance 1995work on (NIMD2004p.14). party respected Latin America,define thestructure of competition in ifand,a state, itis a stable moresystem, promote Mainwaringlawsexisting and electoral (NIMD 2004 rules” p.14). They further suggest party that systems itwith party,interacts and where another more, political at least accordingwith often to and Scully largerforms partof a whole, is, “eachparty That system. provide party of the ‘parts’ that the define themselves is it parties the forDemocracy Multiparty party Institute Netherlands For the developed. systems a years party In recenthaswhat constitutes synthetic amore definition of system been “as the set aparty we adequately system? define of parties…obscures more it than illuminates” (1990 p.316). If not numerically, how then can when argued, analysing party systems “ajudgementas to numberthe of major (See Mair9 for amore 1997Chapter detailed of discussion As Sartori famously this topic). literature unable to profusion of terms” (Macridis 1967quotedin dealSartori 1990 p.317)that have emerged in the with the subtle theorists have endeavoured to movedifferences awayfrom this definition, finding “confusionthe and between Inrecentyears,democracies mainland1954). (Duverger however,and European other party systems acrossby France characterised characterisedmulti-party bytheUK governments, US,and and the the globe traditionally governments, between two-party contrasting normally competition, electoral party systems have been classified by simply counting thenumber of parties in partaking enigmatic elementin study the of contemporary democracy 2006p.566).(Lewis Traditionally, Svassand 2002). As Lewis notes, “party systems have been singled out as being a particularly particularly, a more and party system what exactly constitutes about much hasdebate been There howWhat isa 2.1 Party System? to clearly2 Party Systems distinguish party systems from political parties (Randall and 4 CEU eTD Collection the rulesthe of game’,the potentially democraticundermining legitimacy (1995p.22). more likely are ‘violate by andtherefore less constrained to parties actors political systems, party due ininchoate Mainwaringand suggestthat tothis Scully emphasis on personalism candidates more likely to pursue crowd-pleasing measures over policy enduring Finally,goals. (Tavits 2005p.238). Furthermore, inchoate party systems personalism, encourage with make unable to long-term the necessary developmentforpolicy continuous commitments Scully Additionally,1995 p.25). supply the on fluctuatingside, meansparties support that are their for account actionsmany over to years and several elections& (Rose Mishler 2010p.802;Mainwaringthem & holding parties, with relationships long-term form to volatility citizens allow electoral of level low their with systems party well-institutionalised electorate, the party inchoate systems in pursue which actors their own short-term objectives unpunishedby isunlike an institutionalised accountability.that and suggested in party democratic system It party system suggested in literature,most the notable the being correlation strong the between a well-institutionalised to benefits several are There consolidation. for democratic criterion ifimportant not essential for a requirement functioning democracy, andakey measurement As suggested in my introductory remarks, well-institutionalised party systems are an 2.2 Whydo they Matter? my analysis. throughout be employed will it for(Mair government Given 2007p.3). wideapplicationthe definitionthis inof the literature, when trying to make sense of those structures of competition is to focus on the competition example, party systems are constituted by theirstructures of competition, and the best strategy 1997 p.203) that has in (Sartori ofMair between parties competition importantly comepatterns andmost interactions, to define party systems interactions in patterned competition the parties”(1995among p.4). It is this ideaof within the literature. For Peter Mair for 5 CEU eTD Collection they Mair they on work, predictability.his ahighimportance that place earlier hassuggested In become systems become institutionalised “patternswhen of interaction among political parties andnot merely as electoral vehicles for personal success. suggest Mairand Bértoa party that is, (1995p.5), leaders independently that areconsidered party of leadersorganisations their party organisations play important rolean andarenot marginalised tothewhims of ambitious limited” (1995p.4).Additionally, they withinsuggest that institutionalised party systems, an institutionalised party Wheresystem. such stability exist,notdoes institutionalisation is characteristicis regularly as which majorappearandthenjust of parties not quickly evaporate suggesting that, “patterns of party competition must manifest some regularity…A venue in institutionalisation, system party of measurement important most the is stability that argue also the key measure of the level of party system institutionalisation. Mainwaring and Scully In keeping with Huntington’s 1968 definition presented above, stability in sensea broad is 2.4 Party System Institutionalisation paper. this in employed be will that institutionalisation of definition Hicken and Kuhonta disciplinethe including Mainwaring andTorcal Levitsky(2006), Birnirand(1998), (2005), (2011).unpredictable manner. This definition has been effectively employed by many scholars within Given its(1968 p.12); is,that aninstitution isthat established firmly and unlikely changeto in an parsimony itis seen as ‘stable’ when institutionalised becomes aninstitution that definition Huntington’s and salience makes usep.4). The vast of institutionalisation majority concerning literature the of in the literature,institutionalization when we areable todefine whatitis thathas been institutionalised” (2010 it “wecan AsCasalargues, only when wereferspeak of Bértoa meant to‘institutionalisation’. is the inFirstly, order to measure levels andcauses institutionalisation, of wemustdefineis what 2.3 Whatis Institutionalisation? predictable and stable over time (2010 p.3). Here we can see that along with stability, 6 CEU eTD Collection in somewhat conflicting findings. Traditionally, those factors factors consideredinfindings.conflicting those somewhat Traditionally, toaffectthat party andimportant insightful intostudies this causal mechanism. Unfortunately,have they resulted of number growing a been have there increasingly that however, note We should overlooked. if more indepth study of causalthe mechanism behind this hasphenomenon been mistakenly Given2007 p.156). importancethe of a well-institutionalisedit party wouldsystem, appearas & Zoco Mainwaring p.1; 2010 Bertoa (Casal similar ways other in many are that countries system institutionalisation and why we are able to observe such high levels of variation across what influenceslack of party into hasbeenadistinctresearch above,there As mentioned 2.5 Causesof Party SystemInstitutionalism– The Stateofthe Art institutionalised while some willbecome lessso(1995p.6). more become will party systems some is, progresses that astime process, linear rather be seen as a continuum.As Mainwaring and Scully argue, institutionalisation is nota and party distinguishes institutionalised inchoate between variable that butshould systems, Finally,itis important tostress partythat system institutionalisation is not adichotomous some regularity. manifests andstable understanding that indicates aparty system is institutionalised when party iscompetition a party system.For these reasons,in this paper Iwill employ the Mainwaring and Scully it tosuggestthat we data, would seemincongruous can analyse predictability the therefore of same parties? Furthermore, due to the fact that both Mair’s and my own analyses rely on past for vote the a consistently of number citizens similarelections, do and contest repeatedly to form lasting bonds with parties is in itself an indicator of stability – i.e. do the same parties stability and predictability are one and the same phenomenon. The fact that citizens are able of aggregating and representing citizens’interests. Ihowever would suggestinthat casethis parties, with lastingandform bonds particular without predictability, citizens are unabletocrystallise their identities (1997p.182) political 7 sine quanon if parties are to perform their task CEU eTD Collection changes in the economy and poor levels performance, for economic accountable their hold governments retrospectively voters fact that of economic development will be with systeminstability institutionalisation. In terms of ineconomic factors, party evidence of levels suggestsinfluence to found been that due also have to the sentiment religious of levels or education level as of such variables person-level and socio-structural economic, studies In some 2011 p.8). in part” (Bartolini Hooghe and 2003 quoted Dassonneville take Birch voters and Mair 1990; stable therefore and partisans strong the only turnout… lower a with elections “in that fact the withinlevels of turnout in resulta country higher to levels of party system instability to due is, higher that expect opposite, the to bemay reason also there hand, the other on However, institutionalisation. of levels low signifying therefore citizens, with engage effectively traditionally arguinglow levels that of electoral suggest turnout parties that arefailing to with theorists in literature somethe has debate been a causeof turnout Voter parties. between due Hooghe 2011p.8)tothe overlappingpolicyinevitable of space and the similarities and (Dassonneville institutionalisation system party of levels with correlated positively be will an system numberthe electoral within ofparties arguedthat have far backas Sartori as theorists system party Secondly, institutionalisation. system of party powerful determinant isbe most often assumed to the by region) (often geographical transition since democratic platforms (Ferree 2010p.764; Rose & Mishler 2010p.802). Given logic,this length of time population as they struggle toorganiseinto themselves movementsrobust with cogent transition, parties are notoriously weak and unable tofoster more than superficial ties with the seen as inevitablean growing pain In democracies. of yearsthe following democratic andRoberts Wibbels 1999; Tavits 2005, 2008).Simply,low levels institutionalisation of are political arenas Mair(Bartolini 2003; and al.Birch 1990; Dalton et Lachat2007; 2000; and institutional the within found predominantly been have institutionalisation system 8 CEU eTD Collection that possessthat large indigenous populations more likely to witness poorly institutionalized party in(2005), most importantthe determinant Latin America is ethnic countrieswith composition group countriesor with multiple nested majority (p.778).groups Similarly,for Madrid only one, majority ethnic group having less volatility than those countries with no majority institutionalisation is ethnic homogeneity of a country, with those countries with one, and African states finds that the most powerful factor in determining levels of party system presidential) were the strongest determinants. Conversely, Ferree’s 2010 analysis of 36 party institutionalisation, party fragmentation, and type of regime (parliamentary vs. semi- (sample includes Czechthe Republic, Hungary, Poland and findingSlovakia) that the level of causesof in investigated party Centralinstitutionalisation system East democracies European the level of party system institutionalisation. Casal Bértoa (2010) for instance, has Recently, anew group of independent variables has been found tohave significant effects on change election dramatically a country's(2011p.12). aggregate results” could preferences, which may havedifferent newvoters in these adramatic change turnout, or and Dassonneville comment,Hooghe “if,for example, is there a newly franchised population, As arena. electoral inthe unpredictability of levels high with institutionalisation system party on could anegativeeffect have electorate of the in alternations size the that found notably 2011 p.9).Finally, similar tothe argumentin favour of Bartoliniturnout, and Mair (1990) needthe for parties toprovide cognitive shortcuts when voting (Dassonneville andHooghe mobilization’ of process ‘cognitive the due to institutionalisation levels tolower system contribute of party education and growth the massof media (particularly are alsotelevision), thought to andWibbels in 1999,Tavits 2005 quoted Neff Powell & Tucker 2009p.18).High levels of (Roberts institutionalisation system levels of lowerparty andconsequently electoral arena the (Dalton (Dalton 1984; 1970; InglehartShiveley 1979) and the consequent reduction in 9 CEU eTD Collection structural variablesstructural and analysed. person-level socio- economic, and institutional and political both with institutionalisation system party My analysis will considerboth the most recent findings and the more traditional causes of and was introduced societal the parties playedin role (p.155). this period democratic transition that is the key determinant, but rather the period in which democracy Finally, Mainwaringand Zoco(2007)importantly argue it isthat not length the of time since systems asindigenous citizens repeatedly theirshift vote to unstable, new parties (p.1). 10 CEU eTD Collection an intensive analysis analysis an intensive of a small number of cases we will bemore ableto,“answer likely those 1990; King, Keohane, and Verbain1994 quoted Liberman 2005 p.435).Similarly in applying andsystematic procedures inattention to rival explanations (Achenand Snidal Geddes 1989; of few cases,(Lieberman Unlike2005 p.435). aresearch design based solely anintensivearound analysis a mixed methodscomparative approach making the improves of in valid formsprospects inferences causal of and other cross-national research avoids by strategy analysis) “thisintegrated As (nested science researchis argues, Lieberman present. drawingthe problems on This will ensure much-needed thatthe essentialthe analytical rigour in contemporary social distinctof selection section.for themore2005p.435) therelevantcases detailed(Lieberman justify case-study strengths bias, select weareableandbetter to tousethestatistical Furthermore, analysis approach. statistical of lacktwo abroader important of those with of analysis case-study intensive both of advantages the synthesise approaches”itis terms, to In simple possible approach. analysis nested to the several There are advantages play in party determining system institutionalisation. path can dependency more role that closely the causes andconsider significant statistically the cases from divergent statistical the This will analysis. a morenuanced provide of examination regression. Following this, Iwill an provide indepth case study focussing on one of the multi-level and regression linear multivariate bivariate, run also will I institutionalisation. variable dependent that will volatility) (electoral asaproxyserve for party system in asplaying literature the a causal in role my with party institutionalisation) system have independent variables (those been that suggested avariety analysis (LNA)correlating of Verba (1994),and Collier (2004)amongst Iwill others. firstutilise alarge-Nstatistical by as advocated Lieberman (2005), ‘mixed-methods’ My will approach a employ research 3.1 Mixed-Methods Design, Data, and Hypotheses 11 CEU eTD Collection four) indicators are of completely equal importance. Finally, a distinct weakness of Mair the weakness of adistinct Finally, importance. equal completely areof indicators four) remainsindicators Mair unclear with seemingly suggesting three(and that the sometimes concept, ornot (Enyedi & Casal 2010 Bértoa Additionally,p.11). weighting the of the a unidimensional iswhetherspecify they that relatetoeachother, indicators four) (sometimes theMairapproach. with drawbacksIt remainsunclear associated astohowthe certain three Therearehowever (Lieberman 2005 pp.435-436). country andcomparisons cross-region employs numerical indicators that allow for more rigorous, logical and systematic cross- government(Mair and2010 Bértoa p.3). One of the key strengths of operationalisationthis it to access and in government innovation in government, alternation change, of Frequency system: areas of party the distinct four) on (sometimes three focuses operationalisation O´Dywer (2004);andRybá number of scholarsincluding Toole (2000);Linz (2001);Müller and Fallend (2004); been well within received having discipline, the been in employed various forms by a large 2007) hasalsosuggested by Mair,(2001, An Peter that 10). operationalisation, alternative society (measured bycomparingvote presidential compared lower-chamber (pp.6- with seats) index in party inter-party via Pedersen’s volatility)of competition (measured strong and roots institutionalisationvariables: example party of system two through analysed regular patterns Democratic Institutions: Party Systems in Latin America developed in recent yearswithin political science no with unifying methodbeing established andfewinnovations being(Casal Bértoa & Enyedi The operationalisation of system party institutionalism2010be hasproven acontentiousto topic p.6). In their3.2 Operationalisation classic work confidently (Liebermaninferred” 2005 p.440). assess orbecauseof data could statistical thenature to be causal the relationships not order leftquestions by opened –eitherinsufficient LNA therewere analysis) because the (Large-N Ĝ (2004) (in Casal Bértoa & (2004)(in Casal EnyediBértoa Mair’s2010 p.15). 12 (1995), Mainwaring andScully for Building CEU eTD Collection Sartori Sartori refers toas ‘mechanics’the of systemthe (Mair 1997p.206). We would therefore what for government”, competition especially the and inter-party competition, of “structure system. As weshoulda party how party awell-institutionalised furthermore, refer to define system, and above, dependentvariable in it study,this is whenworth again specifying exactly meanwhat we when we the as volatility analysingelectoral employing for justifications theoretical the considering Before party systems mechanical.theoretical and wecategories: distinct two into fall justifications areThese justification. of amount a certain requires explicitly& Casal Bértoa concerned Forthcoming) 2008; Enyedi (Tavits institutionalisation systems criterion party measurementof appropriate and therefore withthe decisionThere hasbeen much within as volatility is to whetherdebate electoral literature the an theto focus on solely on this variable volatility for thecountries within thedataset. electoral trends particular, for competition government. In order tomeasure phenomenon,will this I consider 2007 p.156),regimes develop(Mainwaringof stablepatterns in party competition andZoco which degreeto institutionalisation: the aspectof democraticone political party system comprehensive data, and the large scopeof given the institutionalisation, this in research, adequately difficulties acquiring number of cases that willAlthough be analysed I will Variable 3.3 Dependent focus on just there are severalfor a more detailed discussion on this topic). strengths party itssystem reliance due to variables ondichotomous 2010 (seeEnyedi & Casal Bértoa approach is thatin it’s original form at least;itdoes not provide anuanced understanding of associateda with Mair’s schema for party system 13 CEU eTD Collection amongst haveothers all insightful produced and convincing analyses, avoiding inevitable the Dassonneville (2011),Neffand Hooghe Powell andand Tucker(2009) Drummond (2006) institutionalisation.Mainwaring (2007),Erik Ferree Zoco and (2010) Lane(2008), previously as aproxy volatility variableused electoral effectively levels for system of party Dassonneville and Hooghe 2011 p.6). Additionally, several scientistspolitical have insight overtrends on time and betweencountries”differences (Lane and Ersson2007in parsimony Pedersenthe of Index is much anditappreciated is therefore often used to give theoretically meaningful way (Pedersen1979p.4).As Dassonneville argue,“theand Hooghe for and possibility easy interpretation allowing the casesin for comparing contrasting and a for each election anumerical score This indexprovides Pedersen Volatility (1979). of Index uniform consensus across disciplinethe onhow itshould be measuredusing thewell known isnearis for theuse there that volatility of justification mechanicalelectoral The strongest label”party 1998 p.590), further (Toka undermining stable patterns of competition. volatility suggesting is that there “no payoff positive sticking toacurrently on unpopular provide information on elite for support party democracy, with levelshigh of electoral parties” Similarly, 2008 (Tavits p.540). supply the on side, analysis volatility of electoral will that large partsoccurring. As Tavits argues,“a highlevel electoral of volatility, indicates argumentgoes, the of the electorate belevels of will volatility able electoral to examine isin whetherthis phenomenon fact haveparties’ presentleaders, issue positions, record and1998 p.590). Measuring scandals”(Toka not developed of andsupport establish labels party “have that some valueindependent from appealthe the of loyalty stability partiesis that elections repeatedly andcontest arethereby to buildable bases stable towards anygovernmentis stable As ornot. discussed above,theminimum criteria party for system of thefor competition the extent what to consider to existing institutionalisation system party expect 14 CEU eTD Collection (adapted from (adapted 1998 p.590).Toka out =20% goes business. beof total will volatility therefore (|60-40|+|40-40|+|20-0|)/2 The vote. At the subsequent election Party A receives 60% receivesof the vote, Party40% ofB 40%,the and vote, Party PartyC B receives Party (A,A B, and contesting there are suppose parties that three C)anelection. For example, 40% of the vote and Party C receives 20% of the receivedvotes by party in that periods tand time t +1(Neff Powell and 2009p.1). Taylor The formula is as follows where n isSystems: ChangingPatterns ofElectoralVolatility. number of parties and pi Volatility originally inof published his path-breaking 1979workrepresentsIndex Pedersen known widely the theemploy will I percentagevolatility measure to order In of Measurement 3.4 2006; Granberg and Holmberg 1990; andLane 2007). Ersson functioningin of democracy to stave order off political play, must parties level acertain tobe forvolatility is considered necessary theproperof that role essential the undermine undoubtedly volatility of levels high While democracy. anegativenecessarily for is not some thatvolatility however, tonote electoral It is important understanding of phenomenon.the findingscomparable with scholars of forother previously amore allowing nuanced Furthermore, this meansalso that any findings from resultthat study this will be directly phenomenon. aparticular of indicators several weighing when emerge that complications 15 immobilisme The Dynamics of European Party (Toka (Toka 1998; Drummond CEU eTD Collection 2 1 Depression Worldof War 1930sor the II” (Pedersen in1979 Dassonnevilleand 2011 Hooghe analysing underlying and causalfactors, “the not impact of disruptive like events the will consider the vote sharescontain results solely from parliamentary elections. In inthe case of a two-chamber theparliament, I lower house only. electoral volatility (Dassonneville 2011p.13;Lijphart andHooghe 1994 p.3), theanalysis will Additionally, in order on ensure of with In order literature inprevious comparability subject to the keeping and to ensure I am theperiod underanalysis.higher throughout Middle East and Africa Central Europe, Eastern Europe, North and South America, South Asia, Australasia, the economic The development. datasetincludes a range of countries from Western Europe, and with democracy history, experience of in are strikingly terms diverse in dataset the contained countries The elections. 633 and countries in 68 volatility electoral on information Volatility: Evidence Postcommunistfrom Countries NeffEleanor Powell and Joshua Aaron Tucker’s (2009) Stabilization of InterpartyCompetition, for his created Zoco, alongside 2007 paper written Edurne two distinct sources. Firstly,from the work of Scott Mainwaring,in particular the dataset The electoral volatility figures thatmake up dependentthe variable in study this come from 3.5 Dataset poorly institutionalised party systems. or inchoate have to considered volatility electoral of levels fluctuating or high experiencing degree ofinstitutionalisation” (Rose& 2010 Mishler p.808) with party those systems the greater the next, the to election one from changes vote party’s each less “the simply, Put Please see appendix for full list of countries included in the dataset inthe included countries listof full seeappendixfor Please For the complete dataset please contact 2 . All countries within the dataset received Polity scores of at least2 or [email protected] and secondly from the from the dataset created for 16 1 . The completed dataset contains dataset The completed New Approaches to Electoral Political Sequences andthe CEU eTD Collection democratised earlier, partiesdemocratised earlier, played amore defined rolemobilising in and citizens society, that countries in those that holds which thesis effect’ ‘period the support findings volatility moreregimesthan inaugurated recently(Mainwaring & Zoco 2007 p.156).These in have that regimes democratised periods earlier lower much electoral competitive volatility, evidence that even when controlling for several other factors though to affect electoral democratisation took place (2007 p.155). tookplacep.155). democratisation (2007 In paper 2007 their measuringlevels of party system institutionalisation, butrather periodthe in which when factor thatis by electorate thedetermining experienced an number of elections or democratisation since time of length not is it that contend however, Zoco, and Mainwaring party and the institutionalised. volatility system beseen will decline as more will electoral increases, of as number elections the that expect We therefore would option. term electoral able increases, citizens arebetter elections todetermine provide which arealisticparties long- numberas by experienced arguingelections of (Converse numberthe that 1969), country of compared have also scholars Similarly, institutionalisation. of system party determinant be inmay a powerful acountry well election democratic first number the years of since that politics andform withdurable relationships particular parties. We may hypothesise therefore electoral to acclimatise electorate as the levels ‘normal’ settle to time to takes that something is volatility that being logic simple The democracies. newer than volatility of levels lower experience will of second-waves democratisation, firstor ineither the elections democratic firsttheir witnessed is, that countries those that older isthat democracies, discipline As presented above, althoughithas been much debated, generalthe consensus within the 3.6.1 Volatility and Time VariablesandHypotheses 3.6 Independent where available. onwards befrom democracies)1945 will oldest employ the earliest Iwill the (in analysis p.10), 17 , the authorsthe found convincing CEU eTD Collection contains those countries that democratised between 1974-1989). There 19countries countries are thatdemocratised between 1974-1989). those contains that the purpose (for democracy paper,thisof this 1974 group after countries to thattransitioned 3 – Group This is group made up of Huntington’s ‘ThirdWave’ isdemocracies, that those Tobago. Israel, Italy, Jamaica, Japan, Malaysia, Netherlands, Norway, Sri Lanka and Trinidad and Luxembourg, Malta, Austria, Belgium, Botswana, Costa Rica, Denmark, France, Germany, consistently from guaranteed 1943 onwards. Thereare 18countries thatmake upgroup 2: democracy (e.g. Germany pre-WW2), however, democratic elections have only been during theperiod 1943- 1973. Several of countriesthese had previous experiences with 2 – democratisedGroup includesThis countries - Wave that group those Second democracies Australia, Canada, Finland, Ireland, New Zealand, Sweden, Switzerland and the USA. governance since this time. Thereare10countries included 1:Great in Britain,group Iceland, Huntington’s ‘First Wave’ and(1828-1942) have consistently maintained democratic 1- OldGroup democracies -This includes countries those democratised that during Democracy’ (1991 p.14): groupsstudiedthree according byMaiwaring andZoco) toSamuel ‘Waves of Huntington’s further study. Consequently,mywill split four I distinctsample into groups (as opposed to the justifies it believe I literature in the salience its and argument this of strength the Given (Mainwaring2007 p. & 166). Zoco society, failing todevelop strong the identities that they didin early democracies in laterdemocratised periods parties have a played significantly less in important role in several generations. and years many over allegiances in strong resulted in turn which networks, Furthermore,social creating the ‘period effect’ thesis holds that in those countries that 18 CEU eTD Collection could explainpolitical attachmentdeepening to parties (in & MainwaringZoco 2007p.161) citizens’ elections exposureto (1969) arguedthat In hisincreases. Converse article, classic is within thenumberof will support literature that volatility gathered reduce as elections the A further in argument, some ways similar totheabove effect’ hypothesis, ‘period has that be: will consider will I variable independent first the this, Given more a be will findingrobust if we canshow this this to bethecase over 4‘waves’ as to 3. opposed volatility, electoral of levels in role determining a plays democratisation institutionalisedifferentspeeds. at Furthermore, ifweconclude are to period that of they will hypothesise that we may todemocratic prior transition political parties Firstly, given the vastly different experiences the countries within the two groups had with 3and4separately. consider groups to theoretical reasons strong are in casethere I believe this Montenegro, Poland, Romania, Russia, Serbia, Slovakia, Slovenia and the Ukraine. Czech Republic, Estonia, Georgia, Hungary, Latvia, Lithuania, Macedonia, Moldova, countries thatmake 4: upgroup Albania,Bosnia-Herzegovina, Armenia, Croatia, Bulgaria, USSR orformer satellite that states transitioned democracy to during 1990s.Thereare21 the Group 4 – Post 1990 democracies- those countries that were formerly either members of the Spain, Turkey, Uruguay and Venezuela. El Salvador, Greece, Honduras, India, Mauritius, Mexico, Papua New Guinea, Portugal, make up group 3: Argentina, Bolivia, Brazil, Chile, Colombia, Dominican Republic, Ecuador, 1) positive correlation in the data if group 1 is coded as 1, group 2 as 2 and so forth. a observe to expect would we then correct, indeed is volatility of levels lower have Period ofdemocratisation –If the hypothesis that elections in older democracies will 19 CEU eTD Collection Given this, the third independent variable I will consider will be as follows: be as will consider will I variable independent third the this, Given captia GDPwas able accountforto 60.6% of variance involatility scores (significantat .000). determinant of electoral volatility Mainwaring with suggesting and per (2006) Torcal that be astrong to development economic shown haveseveral previously studies Furthermore, support new voters encouragingelectoral &Wibbelsto (in alternatives” Roberts 1999p.577). “undermining established political loyalties, increasing anti-incumbentvoting, and measured(often therefore by can GDP PPP), beincrease assumed to volatility by, not)managementaccording totheir(or successful economy.the of hardship Economic and (1978) have Tufte suggested thatvotersretrospectively orrewardincumbentspunish witnessingmore highly institutionalised Kramerparty systems. (1971); Lewis-Beck (1988) performance economic of levels higher with countries those with institutionalisation of party system been as adeterminant seen has alsotraditionally performance Economic 3.6.2 Volatility and the Economy will variablebe: in analysis therefore the The secondindependent to form strong and consistent relationships with a significant number of citizens. of number asufficientelections contested have parties where democracies older than volatile more be weshouldthem to expect innew democracies, experienced of number elections jumplow for them likelihood and between duetothe Consequently, decreasing to the parties. 2) vice versa. and volatility of levels high have will elections few experienced have that countries the hypothesis If volatility. average country’s with the dataset the forwithin eachcountry recorded is correct,Number of elections we would expect to see a negative correlation, that is, – this variable will simply correlate the number of numberelections the of –will variable correlate simply this 20 CEU eTD Collection political power is fragmented or non-fragmented, dispersed or concentrated” (Sartori dispersed (Sartori non-fragmented, political orconcentrated” isfragmented 1990 or power feature albeitroughly,important indicates, an of politicalthe system: extent the towhich immediately “the of number parties that, famously contending with Sartori literature has within beenof asubject longparties party competingaparty system debate within of of number the andparticularly fragmentation, of study The parties. between interactions of number a small for allowing fragmentation of levels low with institutionalisation measure)Parties hasalsobeen foundplay to role inlevelsa determining of party system and fragmentation of Number Taagepera’s using Laakso Levels (measured Effective of 3.6.3 Volatility and Fragmentation 3) volatility decreases. volatility electoral increases PPP isas capita GDP per here, that see anegative correlation management by the government. Given the findings of previous research I expect to to be a retrospective action; that is, ajudgement based on previous economic year$1000 fortheprevious fact duetothe election economicisvoting that suggested contains information for 407 of 633 elections. The variable will consider the GDP PPP Montenegro (from 2001) andSerbia (from 2003 onwards). datasettherefore The The data is available for all countries within the dataset form 1980 onwards except for the World Bank website ( is from data The available countries. between comparison it directallows for fact that has in the United States. The strength of this indicator is its easy availability and the as dollar GDP powerover U.S. samepurchasing the international havingthe dollar with parity power rates an international usingpurchasing converted to dollars dataset for the previous election year. PPP GDP considers gross domestic product variable will GDP percapitaconsider the PPP($1000)for in figure each country my GDP per capita PPP previous election International year, (Current $1000) http://data.worldbank.org/indicator/NY.GDP.PCAP.PP.CD 21 –this ). CEU eTD Collection number of number and of parties, viis votes proportion the of of ith the party. formula andby N= provided Laaksoo Taagepera(1979): 1/ 3 follows: as be will therefore analyse will I variable fourth The fromswitch one party anotherto (Mainwaring& 2007 Zoco p.162). more likely voters ideological inspace with to therefore overlap or of programmatic terms hand,other in party those systems with many parties competing thereis thought beto greater On the institutionalisation. system party of levels high and volatility electoral of levels low which in in resulting (Mair 1997p.200) and consistently present arewell-known electoral alternatives competition centripetal moderate towards tend systems multi-party fragmented Thep.317). dominantstrandin literaturesuggeststhe twothat party systems and theleast The measure of the effective number of parties in a given election is calculated using the 4) http://www.tcd.ie/Political_Science/staff/michael_gallagher/ElSystems/Docts/Election Indicesat datasetwhichisavailable as an following the dataset online address: majority of the countries in my dataset from Professor Michael Gallagher’s Election Indices.pdf (ENP) Level of fragmentation previous election. Given the above, I would expect to of product the were the that of parties number effective see the and volatility electoral a positive correlation in the between relationship will the I consider phenomenon, aretrospective fragmentation, of levels high to response be a to seen are volatility electoral in changes fact the that of633 elections. Dueto for 550 information contains dataset therefore 2000, Montenegro from 2002, Serbia from 2000 and Ukrainethe from 2002. The Herzegovina andGeorgia. Data is available for Albaniafrom from 2001, Croatia provides figuresENP for all countries in my dataset save for Armenia, Bosnia- distance from a puretwo-party included 1987p.410).The data system (Jackman here 3 for the elections under analysis. Levels of ENP are available for the vast the for available are ENP of Levels analysis. under elections the for . Put simply, the higher Effective Number of Parties, the greater the – Here I will consider the Effective Number of Parties 22 Ȉ v2i v2i , whereNis theeffective CEU eTD Collection Consequently, my variable Consequently, fifth be:will parties inwill act unison (2010 p.247). political of branches legislative the and executive the democracies parliamentary within that On and Shugart). otherhand,the argue, asSamuelsandShugart one would naturally assume destabilizing party the increasingsystem andvolatility electoral in 1958 (Linz Mainwaring likelymore govern in to mannera populist and undermine party programmes potentially dependent on a party for financial and electoral Consequently,support. individuals these are beless to aparty less committed and platform likely are to outsiders These chief executive. that presidential systems are more susceptible to political ‘outsiders’ winning the office of frame (Linz in1985 Mainwaring and Shugart1997 p.143). Additionally,ithas been argued much in tootime the accomplish totry tempted to often and therefore are their projects pursue to the set term limits within presidential regimes, presidents are given relatively little time to higher levels of electoral volatility and lower levels of party system institutionalisation. Due regimes (Linz 1985in Mainwaring and1997 p.141)and Shugart consequently inwill result democratic stable sustain to parliamentarism than likely less is presidentialism that suggests There areseveral for thisreasons in conclusion. Juan Firstly, his 1985work, Linz famously thepredictability systems. party the undermine system and stability of than parliamentary system with semi-presidential and presidential systems thought of as more likely to Regime type isalso thoughttoplay important rolean in creating awell-institutionalised party 3.6.4 Volatilityand Regime Type 5) divided into three categories: Parliamentary, Semi-presidential, Presidential according Presidential Semi-presidential, Parliamentary, categories: into three divided Type ofregime institutionalisation). system party of levels (high volatility electoral of levels low with correlate will fragmentation of levels low is, that data, –HereIwill the power consider of each head of Thisstate. will be 23 CEU eTD Collection presidential as 2,and presidential as3. regimes correlation light of theoreticalthe background presented above, I would apositive expect to betweenexample the of thisvariables sort in the parliamentary majority” (Samuels 2010 & p.28).AsShugart Switzerland is only the dataset it isif not includedparliamentary but is which emergesfrom legislature the to notaccountable the “executive the, when analysing regime Finally,be in mustregimes considered dataset. asaspecial the Switzerland case in type. Inregimesin There(Duverger 2010 p.5). are19semi-presidential 19080 Samuels &Shugart are coded governmentas 1, and whosemi- is, along withis is election who headprime who minister heada presidentof state…a of of the cabinet, responsible“the separate types: regime both previous of the with similarities share regimes to the assembly majority are 16examples of presidential withinregimes dataset.Thirdly, the semi-presidential branch…the executive’s term in office is fixed” (Samuels &Shugart 2010p.4). There remove anassembly majority headuniversal suffrage… cannot executive the the of executivethe andlegislative branches of –usuallygovernment directthrough both elect separately “citizens in which, is one hand other the on regime presidential p.4). There are32examples A of dataset. within parliamentary the regimes 2010 between &Shugart (Samuels elections” the may executive andreplace remove majority aparliamentary which by mechanism, confidence the through parliament to responsible collectively are who cabinet and minister aprime of consists branch executive “the inwhich is one regime a parliamentary operationalisation, Shutgart’s to setavailable at (Data Ministers: How the Separation of Powers Affects Party Organisation and Behaviour to Samuels andShugart’s schema within their 2010work http://dss.ucsd.edu/~mshugart/samuels-shugart.html 24 Presidents, Parties, Prime Parties, Presidents, ) . According CEU eTD Collection independent variable I will consider will be as follows: be as will consider will I variable independent sixth the furtherprovide direction, andin some to in order literature the Given debate the volatility. electoral to anincreasein party (Albrightattachments 2008 p.249) anda consequentreduction in educationvery greater few increaseand resulting the exceptions, in political leads knowledge been robustly by challenged Dalton (2007) whohas found strong evidence thatinfact, with logic this has however, Recently, volatility. in and avisible increase is electoral dealignment thoughtful deliberation as information decline”costs (Albright2008 p.248). The consequence loyalty. Put simply, “individuals should be able to abandon rules-of-thumb in favour of more available voters to on an scale,enormous removing importance the of partisanship and party is information meanpolitical internet, the and that television particularly mass media, rely on political parties for direction. Instead, high levels of education and the pervasion of to likely less are individuals well-educated that fact the to due volatility electoral determining Finally I will consider level education.of This is thought toplay importantan role in 3.6.5 Volatility and Education 6) here for download: expectancy by as recorded Unesco’s Global Education Digest 2004(which is available SchoolLife Average Expectancy a more detailed discussion on this topic). Formost countries dataprovidedthe isfor over variables such as school enrolment rates for this reason (see Ram 1998 p.256 for is and time is its for stability over often preferred renowned Life Expectancy (SLE) constant in each country. This would appear to be theoretically justifiable as School as a is itwill be revised fact annually treated that Due tothe not thishowever; measure is available for all countries within the dataset. There are limitations with the data, and countries between comparisons direct for is allows is that variable this of strength http://www.unesco.org/education/docs/EN_GD2004_v2.pdf -Here I will consider average school life 25 ). The ). CEU eTD Collection (1998) andSri Lanka (1994). Malta (2009),GermanyGreece (2007), Mauritius(1997), (2008),PapuaNewGuinea yearthe 2010with followingthe exceptions: Albania Luxembourg(2004), (2008), 26 CEU eTD Collection Lithuania (average volatility in69.28 percent) group 4.In spite of this caveat, averages the phenomenon can be observed with Bolivia in group 3 (average volatility 39.45 per cent) and same average. The by group the explained cannot beaccurately that volatility experience 2, Costagroup Rica (average volatility andMalta (average29.86%) volatility 2.35%) elections where volatility has reached 43.23%(1999),far from group the average. Similarly in In group1for instance, Icelandhas an average volatility at17.29%, of times experiencing all for countries. accurate not, of course, are averages these that however, note, to important markedly average volatility increasingly wemove from as 1 4.group through to group Itis groups 2, 3 and 4 are considered. We are able to see a clear directional relationship with European democracies. The periodisation thesis is further supported when the averages of likely andnot simplyinin more be Western volatility are democracies observed to older at all. What this low level of volatility would more accurately suggestis that low levels of up Group 1 (Canada, New Zealand, Australia and the USA), are in fact not in Western Europe party system (Mair 1993 substantiate Mair’s claim that there is no Western European trend towards ‘defreezing’ of the between 1885and 1985(Mair 1997p.67).Although findingthis would appeartoin part Bartolini and Mair tobe the average volatility in thirteen Western European countries below ten per cent at amere 8.13%, a figure extremely close to the 8.6 per cent found by in Malta in 2003. Within 1, isgroup that theoldest democracies, averagevolatility remains volatility for asingle election being 85.93%in Moldova in 1998 andthelowest being 0.28% The averagevolatility 633 elections in across 68 countriesis with 18.84% highestthe level analyses. andmulti- multivariate my from setbefore moving on to bivariate, data statistics descriptive This section of the paper will consider the findings of my analysis. Firstly I will consider Results andFindings 4 p.81), it is worth considering that four of the 10 countries that make 27 CEU eTD Collection from from 0.28 –66.4% with 19 (8.1%) of 234electionsthe studied seeingvolatility levels of 30%, Icelandin 1999 and in Canada 1993. The second had group volatility scores ranging above levelsvolatility scores 32%were electionsabove of Only observed. experienced two thirteen European countries between 1885 and 1985in noelectionswhich with volatility the range noted here is somewhat wider than the range noted by Bartolini and Mair (1990) for with 1.2% of the elections considered having a volatility level of above 30%. Interestingly, studied.period Group1 (old democracies) had volatility scores ranging from 0.8 –43.23% All fourof during countries experiencedawide groups of scores volatility range the electoral Table.2 4.1 Ranges Table.1 volatility. levelsofelectoral lower the the democratization, across the four groups does seem paint aclear picture: the earlier the period of All countries/All All All countries/All elections Group 4 Group 3 Group 2 Group 1 elections Group 4 Group 3 Group 2 Group 1 – Averageby volatility group – Volatility by ranges group Lower Range Average Volatility Volatility 5.93 0.28 0.28 (%) 0.5 0.8 44.29 21.92 14.47 18.84 8.31 (%) Upper Range Volatility 85.93 71.95 43.23 85.93 66.4 (%) 28 N Countries 21 19 18 10 68 % of Elections above 30% Volatility 19.23 78.3 20.3 8.1 1.2 N Elections N N Elections 143 234 164 633 92 143 234 164 633 92 CEU eTD Collection countries witnessed elections in which the volatility broke the 30 per cent barrier. Finally, 18countriespassedthe threshold the that witnessedin elections and 3,12 of 19 the group elections surpassed 30%threshold the in 2,3 andgroups 4.In group 2for instance, sevenof undoubtedly outliers,far from the norm within the group. On the other hand, many more party,from effectively erased political the electionsmap These are (Johnson 1994 p.1). two saw Conservativethe party, which up until had pointthis been thecountry’s most successful elections competing (Kristinsson 2000 p.187); and 1993Canadian the General whichElection (Kristinsson with2000 p.187) only of only six the two receivingparties in seats 1995 the since to WorldWar” Second the alternatives occur of electoral radical restructuring cent threshold, the Icelandic Parliamentary Election of 1999 which witnessed, “the most relationship. Asmentioned above,only of electionsthe two in 1 exceededgroup the30per we can per cent avery 30again of over see distinct scores volatility experienced that group in of each percentage elections the displays column if the that consider we importantly, 2-4 theuppervolatilitygroups increases markedly.range Finally, arguablymost and the most volatile election falls to 23.26 per cent (Great Britain 1992). As we move through 43.23 percent, if we remove the two outlying cases of Icelandin 1999 andCanada in 1993, The upper volatility range tells a similar story. Although the upper range for group 1 is high at democratised since 1990, it is very difficult topredict the range of volatility at each election. that countries in the Conversely, volatility. electoral of range narrower a experience again see a clear directional relationship. Those countries that democratised earlier have the elections having volatility levels of above 30%. In terms of range-based trends, we can note possible to range widestthe volatilityof – scores 5.93 85.93 with a staggering 78.3%of Finally,in 4, isgroup that thosecountries thattransitioned democracy to init is 1990s, 71.95% with a 20.3% of electionsthe in datasethavingthe avolatility level aboveof 30%. above 30%. In Group 3 the range of electoral volatility scores widened even further to 0.5 – 29 CEU eTD Collection This relationship is clearly visible in graph the below N Countries Table.3 bonds with particular parties. learn partieswhich canbeandwhich supported graduallycan stronger be thrown out, forming across the full sample falls in every subsequent election suggesting that over time, citizens Asvolatility wecansee average below, volatility decreases. increases, electoral in a country of elections number the to this question-as aclearanswer if hereas provides appear data the is literature trends over time. Thatvolatility is, does electoral volatilityelectoral decrease and election on election? It would institutionalisation system party the within theme A key 4.2 Trends over time number of citizens failing to form lasting bonds with particular parties. conclude that the party systems across group 4 have been highly unpredictable with a large Slovenia belowvolatility never experiencing 30%.Given figures,these wecan certainly elections in Estonia, Georgia, Lithuania, Macedonia, Poland, Romania, Russia, Slovakia and every single country in group 4 experienced elections with volatility above 30 per cent with – Averageby number volatility election (%) Volatility Average Election First 30.86 68 (%) Volatility Average Election Second 25.20 68 (%) Volatility Average Election Third 22.97 68 30 (%) Volatility Average Election Fourth 21.64 61 (%) Volatility Average Election Fifth 18.93 42 (%) Volatility Average Election Sixth 15.22 37 CEU eTD Collection suggesting that volatility levels remain unpredictable in countries. these levels volatility unpredictable remain that suggesting 3 isgroup mostunpredictablea thewithout group reductionclear tendency towards Finally, election. fifth the following in volatility rise notable a experiencing before election on election toreduce seeming volatility with is completely electoral clear not relationship the and seventh elections all show increases on the previous levels of volatility. In group 2, again fifth third, overtime,the reduce is to for volatility there atendency 1,although in see group graph below shows the average volatility by numberelection for groups1, 2 and 3. As we can less The becomesaggregate data, somewhat robust. asthepattern to opposed disaggregated maturation byforward as thesis substantiate the if put (2005), consider Tavits we Although this trend seems to be in keeping with some of the literature and goes some way to Figure.1 –Average number volatility by election 31 CEU eTD Collection stable relationships with parties. with relationships stable in the right direction towards a more predictable party system in which voters are able to form at the fifth election remains above 30percent, however,itwould imply they that movingare Kitschelt’s period (2001p.306) trial-and-error toa more settled equilibrium, volatility average election on election.Although may this not suggest thesethat countries have graduatedfrom 4, be volatility however, to clear.In group Electoral relationshipreduces would the appear Figure.2 – Average volatility by election number for Groups 1, 2, and 3 32 CEU eTD Collection more data becomes available. more data maturationthe hypothesis in togroup4 reference countries should bemade tentatively until regarding any conclusions andtherefore less clear somewhat becomes trend the five elections, If we are to consider the volatility scores for only those nine countries that have experienced N 21 Countries Table.4 countries within group. this However, as the table below shows, data for five elections is only available for nine of the 21 Figure.3 – Average volatility by number election group 4 –Average volatility by number election Group 4 55.40 (%) Volatility Average Election First (%) Volatility Average Election Second 21 44.53 21 41.36 (%) Volatility Average Election Third 33 18 37.97 (%) Volatility Average Election Fourth 9 33.93 (%) Volatility Average Election Fifth CEU eTD Collection average volatility scores of 40%. above scores volatility average mostthe volatile 14countries from are 4 withgroup these 14 countries all experiencing average volatility. Inkeeping with patternsthe presented earlier, weunsurprisingly find that It is worth alsoof study to order countriesthe withinform from datasetthe highestlowest to 4.3 Volatility in Order Figure.4 –Average volatility by number election Group 4b 34 CEU eTD Collection experiencing elections ofbelow 10percentvolatility between 1981 and 2005. result,unexpected having one of lowestthe average volatility levels and consistently volatility, Malta, is from group 2 as are Austria and Germany. Honduras is the most countries, the majority are as expected from group 1, however, the country with the least At the other end of the scale the findings are somewhatmore surprising. Of the least volatile Table.5 Position – Highest average volatility by – country average volatility Highest 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Macedonia Lithuania Moldova Romania Slovenia Armenia Slovakia Country Georgia Albania Estonia Poland Russia Serbia Latvia 35 Average Volatility 48.30 42.12 45.25 45.71 46.03 50.47 69.27 45.56 46.92 49.30 55.61 56.64 58.35 47.2 (%) Group 4 4 4 4 4 4 4 4 4 4 4 4 4 4 CEU eTD Collection semi-presidential (Finlandregimes and Ireland)and1unclassified regime (Switzerland). 20 of them are parliamentary democracies with 2 presidential regimes (US and Honduras), 2 investigate further. Similarly, of the 25 countries with the lowest levels of electoral volatility, semi-presidential (15 of 21 countries). This is an area that the regression analysis will countries thatmake 4, and upgroup democracy most recently transitioned to are by largeand asthose variable of herewith period democratisation the Theremay besome(3). conflation with fouronly of 15mostthe volatile being countries either presidential (1) orparliamentary that experience the highest levels of electoral volatility are often semi-presidential systems countries Those type.in regime of pattern terms againbe clear areasonably to There seems Type Regime 4.4 Table.6 Position – Lowest average volatility by country 68 67 66 65 64 63 62 61 60 59 Switzerland Honduras Germany Australia Country Sweden Finland Austria Ireland Malta USA 36 Average Volatility (%) 3.42 6.79 8.03 2.35 6.87 8.04 9.46 6.64 7.44 7.93 Group 1 3 1 2 1 2 1 2 1 1 CEU eTD Collection average volatility volatility scores. average levels of electoral volatility is 6.30 compared with 3.14 for those ten countries with the lowest with those who have lowlevels. The average ENP for the ten countries who have the highest compared volatility levels of haveelectoral countries higher who those between difference fragmentation than those with lower levels of volatility. Itis possible to note a distinct at each election. Those countries with higher levels of volatility also have higher degrees of The level fragmentationof within a party system is considers the Effective Number of Parties 4.6 Fragmentation Figure.5 averagethe life school expectancyis 15.75. 10 most volatile countries is on averageyears 14.45 whilstin 10leastthe volatile countries by the UNESCO Global Education Digest we can observe that school life expectancy in the variablevolatility. asproduced we consider the Life Average If (Years) Expectancy School beThere seems littleto glean in to of terms relationshipthe between and electoral education 4.5 Education – Average school life(years) expectancy –Average school 37 CEU eTD Collection more volatile than those that democratised in earlier periods. in democratised earlier that those than more volatile variables of 0.8374 (p-value countries <0.05).Those democratisedthat inlater period between two the a very high correlation is case.We are ableto observe the This indeed experiencing increasingly levels.higher group 1 would democracies, we would it oldest was hypothesised the positiveHere correlation, that witness a experience the lowest 1 Hypothesis levels of average volatility regression analysis. to movingon before correlation) (Spearman’s analysis with correlation bivariate I will Firstly consider levels significance. they of statistical reach groups 2, 3 and itof is intoensure increases, elections toalso important robustness order consider whether 4 importance of the regarding period and changesof involatility thenumberdemocratisation as interesting aboveAlthough trends particularly the questions, certainly presented raise 4.6 Bivariate Correlation Analysis Figure.6 – Average ENP - Period ofdemocratisation 38 are CEU eTD Collection subject. that experienced below 20% volatility. The regression analysis will shed more light onto this below which shows a huge variation around the line of bestfit, particularly for those elections butrather something more subtle occurringin data.Thisthe is by supported scatter the plot flawed fundamentally is theory the that fact the to due is not this that suggest may correlation theweak although incorrect wasinfact of relationship the direction the Interestingly, The findings. correlation signifiesresult (p-value of 2e-04) 0.2017 aweakpositive correlation. Unlike in hypothesesthe 1 and 2, statisticalthe analysis hereprovides less conclusive volatility. levels high PPPatelectoral yearof election previous will per capita experience the Here the hypothesis suggested a negative correlation, those countries with lowlevels of GDP Hypothesis 3 -GDP per capita PPP previousyear election elections. more have experienced that those than volatility average (p-value < 0.05). Those countries that have experienced few elections have higher levels of Again, this hypothesis can be due substantiated to the negative strong of correlation -0.6048 volatility. hadthat experienced fewer elections wouldlevels experience higher of average electoral aboveThe hypothesis suggestedwewould a witness negative countries correlation. Those Hypothesis 2 - Number of elections 39 CEU eTD Collection 2, presidential as 3 then we should expect to see a positive correlation. The literature wouldif suggestthat parliamentary arecoded regimes as 1,semi-presidential as Hypothesis 5 -Type ofregime volatility. electoral of levels lower 0.05). It would appear as if those countries that have lower levels fragmentationof also have analysis in The bivariate positive (p-value results a reasonably strong correlation 0.4548 < will likely shiftbe more to parties. between systemshighly fragmented limitedwith andideological programmatic meansvoters space For this test, party system theory would suggest that we will observe a positive correlation as Hypothesis 4 -Level of fragmentation Figure.7 – GDP per capita and electoral volatility scatter plot scatter andvolatility –GDPpercapita electoral 40 CEU eTD Collection country and dividing the totalbynumberofobservations anddividing country 4 variable dependent the as country by volatility electoral average with analysis) under period Model 1 focuses on average scores for the various indicators presented above (for the time models linearmultipleemployed. As canbe distinct regression are seenbelowthere two 4.7 MultivariateRegression Analysis The regression analysis will adeeperinvestigation of provide relationship.this lowdue to correlation. the however, tentatively, interpreted be must This volatility. electoral of levels higher have expectancy life school of levels low have that countries Those 0.0023). (p-value of -0.363 correlation negative in weak a reasonably analysis results The bivariate negative. there was not aclear hypothesis over whether the correlation result will be positive or relationship of the direction the uncertainty over andthe literature within debate Given the the 6–AverageHypothesis SchoolLife Expectancy 5. hypothesis substantiates weaknessthe of correlationthe itwould be ill toconcludeadvised that this finding alone Indeed, this is what we see although the correlation is weak at 0.332 (p-value < 0.05). Given volatility scores from separate elections (hypotheses 3 and 4) for useful analysis. forvolatility each (hypotheses 1, 2,5andcountry 6)whereas somehypotheses require with average directly concerned are somefact hypotheses tothe that are necessary due 633 (the total number of elections that we possess electoral volatility data for). Two models volatility for election years as the dependent variable. We therefore have a maximum N of country. Model 2 considers observations for specific elections,years, etc. with electoral for each observation wehave variable N of an 68 –onedependent In model 1 therefore For example, average GDP is calculated by summing the GDP scores in the data set for a particular set fora scoresinthedata GDP summing the by is calculated GDP example,average For 41 4 . CEU eTD Collection . p<0.1; * p < 0.05; **p< 0.01; ***p < 0.001 Dependent isNote: variable average volatility by electoral incountry Model Adjusted R-squared0.69 Model 1 Table.7 Volatility ofElectoral Determinants Life Expectancy Type of Type of Regime Average School democratisation No. of elections of No. Average GDP fragmentation considered Intercept Wave of Average ($1000) (ENP) – Determinants of average electoral volatility – of electoral average Determinants Ref Cat: Semi Pres Pres -11.09 Coefficient Parl Parl -10.93 12.67 -0.73 -0.06 -0.11 2.19 8.51 42 Standard Error 13.72 3.82 3.19 0.81 0.72 1.82 0.06 0.27 Significance *** ** ** ** - - - - CEU eTD Collection model 2 in which the coefficient (7.59) also reaches levels of statistical significance below significance statistical of levels reaches also (7.59) coefficient the in which 2 model in electoral volatility atp (significant < 0.001). This relationship isfurther supported by 1, this isindeed thecase; moving from one group to nextthe produces an increase of 8.51% 1990s), we would see a corresponding increase in electoral volatility. As we can see in model 1, group oldestthe group democracies, through 4,thosecountriesto democratised that in the decreased as of (from 1suggestedthat time period Hypothesis sincethe democratisation Hypothesis 1–Period ofdemocratisation . p<0.1; * p < 0.05; **p< 0.01; ***p < 0.001 for in each 2 Model election isvolatility Dependentvariable electoral Note: Adjusted R-squared0.39 Model 2 Table.8 previous election Fragmentation at Fragmentation Election number Election GDP atprevious Type of Type of Regime democratisation election yearelection Expectancy Life School Intercept Wave of ($1000) (ENP) – Determinants of electoral volatility by single election bysingle volatility electoral of – Determinants Ref Cat: Semi Pres Coefficient Pres -8.16 Parl Parl -8.05 -0.47 -0.21 9.28 0.23 1.57 7.59 43 Standard Error 8.04 0.19 0.42 0.42 1.22 2.40 2.08 0.12 Significance *** *** *** *** - - - . CEU eTD Collection only reached the lowest levels of statistical significance (in only one of the two models), and wewould$10,000 increase see a consequent of 2% in Asvolatility.electoral this variable of forincrease clearly, every More 0.21%. in of volatility we seeareduction electoral significance, albeit the at p < 0.1. For every increase of $1000 dollars at the previous election, other hand, GDP per capita PPP ($1000) for the previous election year, does reach statistical does PPP capita average per GDP GDP PPP for acountry fails to reach any level of statistical significance suggesting that can see above, the results of the two models provide somewhat different findings. Average result in lower levels of electoral volatility, and therefore a more stable party system. As we Here, given the theory it would be expected that higher levels of GDP per capita PPP would Hypothesis 3 – GDP per capita PPP previous electionyear volatility. on effect seem not haveadiscernible does to country each by experienced of elections number the simply, Put variables. relevant the for significance levelsof statistical accepted reaching the the multivariate analysis suggest however, that this is not the case, with neither model to electoral andpolitics developed sturdier relationships with specific parties. Theresults of have could citizens effect the volatility becamemore accustomed of electoral as diminishing HypothesisFor it was suggested2 that number the of elections experienced by a country Hypothesis 2 –Number of elections plays role ahugely important in levels determining of volatility.electoral highly substantively insignificant modelsboth would suggest period that of democratisation is also and significance level statistical highest the of reachesboth variable this fact that electoral volatility and therefore provides substantial support for the period effect thesis. The Substantively 0.001. finding very this showsthe effectlarge of period hasdemocratisation on not influence levels of average electoral volatility. On the 44 CEU eTD Collection presidential regimes causing an increase in volatility and finally presidential regimes causing semi- volatility with levels tohavelowest electoral of the werethought regimes Parliamentary way: following in the levelsaffectvolatility electoral to of Type waspredicted of regime Hypothesis 5 –Type of regime party system institutionalisation. levels of electoral volatility. Clearly, the level of fragmentation does have a visible effect on democracies), wewould seeacorrespondingincrease of 4.71% and6.28% respectively in from say three to six or sevens between two elections (not uncommon, particularly in new increase of 1.57% in electoral volatility. If the number of parties within party system jumped in in election by previous results an of number atthe a party parties system effective the one interesting findings, arguably even more so given its more nuanced analysis. An increase in 13% more average volatility than the two-party system. Model 2 also provides us with observe multi-party should system averages say sixparties, the multi-party system that with a consistentlysystem that two rulingpossesses throughout theperiod parties under analysis volatility. Substantively this is an interesting finding. If we compare a traditional two-party country’s party system increases by one, we see an increase of 2.19% in average electoral coefficient for model 1 shows us that as the average (effective) number of parties in a previousthe election having (p <0.001) effect adeterminate on variable. dependentthe The indeed seems to be the case with both average fragmentation (p < 0.05) and fragmentation at high levels of fragmentation were predicted to cause high levels of electoral volatility. This is, that volatility, electoral on effect tohaveapositive were predicted fragmentation Levels of 4–Level Hypothesis of fragmentation tentatively. interpreted be should institutionalisation system its limited substantive impact,any aboutthe conclusions number of elections causing party 45 CEU eTD Collection that the independent variables included inindependent that variables included for modelsthe two the alarge can account of the part andfor 0.39(adjusted) model both 2 suggest that models arewellfitting. simply, Put given differentmodels are not directly comparable, the two results of 0.69 (adjusted) for model 1 independent variables) is explained by variationthe in X. Although R regarding the goodness of model fit; more specifically, what percentage of variation in Y (the The different R Goodness of Fit negative. is or relationship positive the to whether the relevant statistics for both models 1 and 2, we are not even able to find strong evidence as influence party system institutionalisation. If we construct a 95% confidence interval around school contrary life not literaturemodel average expectancydoes to the suggesting that in either significance statistical toreach failed variable This volatility. average electoral volatility, that is, high levels of average school life expectancy would cause low levels of average electoral on have effect negative to a suggested lifewas expectancy Average school 6–AverageHypothesis SchoolLife Expectancy originally hypothesised. Regime type does seem to influence electoral volatility, although it is not the one that was regimes. presidential and between parliamentary difference significant statistically discernible slightly different coefficients, can also be seen in the second model. Interestingly there is no volatility than parliamentary regimes (p < 0.05); the same causal relationship, although 11.09%produce (p < 0.05)more volatility than presidential and10.93%moreregimes interesting nonetheless. If we considermodel 1, we can see that that semi-presidential regimes but theory the to contradictory aresomewhat results The levels ofvolatility. highest the 2 statistics are also worthy of interpretation as provide us with information 46 2 statistics across two across statistics CEU eTD Collection focuses on within-country comparisons) should be interpreted tentatively. interpreted be should comparisons) within-country on focuses variance in dependentthe variable. Given anyconclusionsthis, drawn from model 2 (which explaining the between country variance, andso not powerful explainingat within country cross-country the varianceresidual suggesting reducedramatically, model the is goodat included model, the wesee are variables When is independent countries. the variance across The results of baselinethe model independent(no variables included) suggest 66% that of the arandom regression model effect’, intercepts has been for to compensate run this. ‘country so-called the may be related, eachcountry within scores volatility the Given that Analysis Regression Multi-level 4.8 independent variablesin are general andwell chosen for appropriate themodel. the wecan that variable, measuringassume in institutionalisation party system variation the 47 CEU eTD Collection that any interpretation of this relationship willbeatbest speculative. of this relationship anyinterpretation that one of the multipletwo models regression failed toreach acceptedthe levels here suggesting GDP per capita PPP at the previous election, which only reached statistical significance in volatility. powerful of mostelectoral the in werealso predictors variables these analysis which multiple of findings the the regression reinforces further andRegime.Election This Type of in the multi-level model,Wave of Democratisation, Level of Fragmentation at the Previous significant remain that variables thereare three that model show multilevel of the The results . p<0.1; * p < 0.05; **p< 0.01; ***p < 0.001 Model 3 –Random Intercepts Model Table.9 previous election Fragmentation at Fragmentation Election number Election GDP atprevious Type of Type of Regime democratisation election yearelection Expectancy Life School Intercept Wave of ($1000) (ENP) – Determinants of electoral volatility random intercepts model intercepts volatility random – of electoral Determinants Ref Cat: Semi Pres Coefficient Pres -8.31 Parl Parl -8.17 -0.74 -0.61 -0.17 8.46 1.74 0.37 48 Standard Error 11.32 0.53 2.40 3.23 1.03 0.65 1.71 0.15 Significance *** *** *** *** - - - - CEU eTD Collection 10% more volatility on average for those countries within the dataset. This finding, that party system institutionalisation, however semi-presidentialism would appeartocausearound The fact of being a parliamentary or presidential regime does not seem to influence levels of contradictory somewhatto the theory. interesting results, 5similarly Hypothesis produced alternative variable as percentagein of such highereducation population beutilised.could infuture, toprovide order amore nuancedperhaps analysis with more data, available an of data available, the figures had to be treated as constants throughout the period of study. In employed. A particular problem with the operationalisation of this variable is that due to lack beand should an variable is alternative incorrect of education be operationalisation the that may also A out. in factpossibility andthird eachother cancel are correct of education effect determining role in levels of electoral volatility or possibly that both theories concerning the multivariate model, wemust conclude therefore, either that levels of education do not play a in reach significance either failed to variables relevant fact the tothe that due Unfortunately, would provide some usefulempirical tosupport findings one of the theoretical arguments. analysis this that hoped was it literature, in the debate the Given parties. with bonds stronger inexpectancy resultlevels lower will volatility of electoral moreformingdue to educated the have recently been scholars who have argued the opposite, that is, higher levels of school life in of importance cognitive previously shortcuts by provided On parties. otherhand,the there levels of will education cause in higherlevelsvolatility turn of electoral reductiondue tothe higher hold that scholars traditionally, brief, In volatility. electoral and education between relationship the of direction the over literature the within debate noteworthy been has there to reach leveleven accepted minimum the previously, Asdiscussed statistical significance. of order, against theoretical expectations hypothesis 6 -Average School Life Expectancy failed reverse In findings. interesting several point towards above presented analyses statistical The 4.9 Discussion 49 CEU eTD Collection semi-presidential regimes (2006 p.220). (2006p.220). semi-presidential The multipleregimes between possible interactions these political lifein dominate and legislature cabinet the the presidency, between the interactions a popularly andelected president government. an assembly-supported Protsyk arguesthat the ‘intra-executive’ coexistence (Protsyk 2006p.219); is, that contemporaneous the existence of this increasein volatility in these with due tothecountries associated inevitable complications now looks somewhat different to that of 50 years previously. The most plausible reason for 1990s (which saw birth the of a number of semi-presidential regime the regimes), landscape was dominated by presidential regimes), and the democratisation of Eastern Europe in the Following the democratisation manyof in countries AmericaLatin during the 1980s (which While the number of democracies by doubled exactly half1983, parliamentary.remained times. AsSamuels innote, andShugart 1950s, the of 20 democracies 12were parliamentary. government (either ‘purely’ presidential or semi-presidential) has risen dramatically in recent is it Firstly, important to note that regimes. in recent years the number of regimespresidential with some form of presidential‘pure’ and parliamentary than volatility electoral more significantly cause would regimes semi-presidential why to as arguments several are There volatility. in electoral weobserveincreases that 2008), forbest corrects the shortcomings of ‘pure’ models (Protsyk 2006; Cheibub &Chernykh is in semi-presidential systems, which has increasingly been seen as the arrangement which system institutionalisation, yet in this case at least, this is not strictly what we find. Instead, it manner (Mainwaring& Scully 1997 p.5).This is hypothesised chances toreduce of party requireare party support more likely come to power to and consequently in govern a populist argued (seeLinz in 1985)that presidentialstrong systems, political ‘outsiders’donot who established theory on the effect of regime type may need to be reconsidered. It has long been semi-presidential regimes do seem to cause higher levels of volatility would suggest the presidential regimes do not seem to cause more volatility than parliamentary regimes, while 50 CEU eTD Collection allegiances more easily. The consequence being that parties moreallegiances easily. differentiate beingthat struggle consequence The to switch in likely allowsto voters which aremoreturn ideological overlaps and programmatic & Roberts Wibbels 1999p.578;Toka 1998). In systems with small distances between parties, incorresponding reduction in distance averageperceived between (Pedersen the parties 1983 spatial of voting. As the number of parties within the party system increases, we see a of theory Pedersen’s employs phenomenon this for explanation salient most The finding? this directly affects stabilisation (Laakso & Taagepera 1979 p.3). How though can we account for from the time of Sartori (1990) have claimed that the number of parties within a system institutionalisation. Indeed, this finding substantiates the claims of party system theorists who models this multivariate that isapowerful suggesting of party system determinant the both in significance statistical of levels reached 4), (hypothesis fragmentation of Level value. be of undoubted would further analysis volatility, electoral and type regime between relationship into the scale influence of regime type; and that this study is, as far as I’m aware, the first on such a large the on literature question intotraditional the call directly models; three over consistent theoretical arguments (Cheibub & Chernykh 2009 p.202).Given myfindingsthat are well-developed nor stability, democratic affect will regimes semi-presidential in which ways the about literature the within consensus no is there that however, note, to important is It holdforsure who governmental decisions to accountable come with presidential regimes, but also asituation in which neither citizens that nor elitesrisks are populist and anti-party potential the only not with stability system party on effect short cabinetlifespans (Cheibub& Chernykh 2009 p.208). This can logically have adirect unstable governments, almost unlimited opportunitiesfor conflict between political elites and of patterns hierarchical (Protysk 2006; control Cheibub & The Chernykh 2009). resultis spheresthree can resultand do in confusion between accountability, and responsibility 51 CEU eTD Collection is, we should not assume that countries with consistently lowlevels of GDP per capita PPP to volatility. That levels causesincreasing electoral PPPforaffects of acountry) per capita that nations (potentially toother measuredbyaverageGDP economic comparative performance 1999 p.577). Given findingsthe itabove, may be that contended itis notsimply poor strife are more likely shiftto their votes to new electoral alternatives (Roberts & Wibbels rational and will for incumbents punish poor economic performance andin times of economic of GDP at the previous election year. Currently, economic voting theory holds that voters are levels to low beseen as ashort-term reaction volatility can electoral ameasure.Instead, crude significantinfluencing role. This averagewould suggest that volatility a country isof too 2, levels of GDP per capita PPP at the previous election were found to have a statistically PPP was found not to have an effect on average levels of electoral volatility, whilstin model the most interesting in this study. This is due to the fact in model 1, average GDP per capita The conclusions we can draw from multivariate analysis relating to hypothesis 3 are among electorate & (Roberts Wibbels 1999p.578). as opposed to exogenous factors such as economic voting orchanges in the makeup of the themselves, systems party effect of endogenous a mechanical, a certain extent are to volatility in changes that isimplies is that notable particularly finding this makes What matters. democracies,in showing numberimportantly, the that of analysingparties post-1990 institutionalisation “the literature, number of matters” parties (Sartori 1990 p.318)andmore countries, adding further even supportto well-knownthe claim in party system fragmentation in influencing party system institutionalisation in such a wide range of first significancethe of tonote the aboveare presented Ibelieve findings the 2009), that (Bartolini and Mair1990; Remner 1991; Mainwaring & Zurco 2007;Neff & Powell Tucker been found tobe cause of asignificant by party systems institutionalisation several studies has this Although citizens. with bonds term long form and opponents their from themselves 52 still CEU eTD Collection so far as volatility stabilises over time, or in line with the number of elections experienced; in institutionalisation hasparty an on system itis that time effect that not conclude therefore must we significance, statistical of levels minimum the reached models in the elections of may well be correct in their analysis. Given that neither of the variables relating to the number and Mainwaring that Zoco suggest analysis andmulti-level multivariate of The results the parties in society during this period. party system institutionalisation, but rather the period of democratisation and the role of Zoco (2007)have itis suggested that not numberthe of determineelections thatlevels of institutionalisedand (Bértoa MairRecently Forthcoming p.85). however, Mainwaring and become for party areneededthe to system multiple andelections time transition, unsurprising and logical in its formulation: given inevitable rupture the in democratic is focus. traditional position The under renewed has recently come institutionalisation substantively significant of findings papergiventhe thethat influence of time on party system It is in the results relating to hypotheses 1 and 2 that provide the most interesting and capita PPPdata (GDPperwas elections for electionsonly post-1980). available investigation into this relationship wouldbe fruitful, possibly considering awider range of tentatively. Given potentiallythe important findings, Iwould suggestfurther that low values of statistical significance reachedin model 2, wemust interpret any conclusions volatility,in this case we are only able to witness a decrease of 0.21%. Furthermore, given the a found in$1000 increase that GDPproducedadecreaseper capita of in 1.29% electoral substantively resultsthe arenot notable. that UnlikeMainwaring who and Torcal (2006) Although theoretically findingthis isinteresting, particularly in terms of theory development, alternatives following short-term economic hardship. for electoral look seemingly voters volatility. Instead, high levels exhibitof electoral 53 CEU eTD Collection This is an interesting finding and would warrantfurther research. no significant correlations with the sixth electoral period, not even the fifth electoral period. are fact there the that dueto cycle electoral sixth atthe tooccur seems unexpected something levels of volatility remain at lowlevels and vice versa. However, itmust be noted that does not affect party system institutionalisation. Instead, those countries that begin with low finding to addingrelatively countries stable within longevity the credence that of regimes dramatically over time, in general, over five election periods at least, volatility levels stay elections 0.410 (p>0.01). While there are some countries that increase or decrease fourth election 0.646 (p>0.01), and even reasonablyremains strong between firstthe andfifth first andsecond election is 0.806(p>0.01),firstand third first election 0.723 (p>0.01), and at the volatility countries’ between set.Aswecan correlation the see, inthedata countries all for number election by levels volatility between correlations shows below table The their societies. and govern views alargedemocracies numberof feel partiescitizens failadequately that their to represent Zoco 2007 p. 166) and Kitschelt have shown,in new democracies, and increasingly in old and governing organisations. However, astheorists such as Schmitter (2001in Mainwaring & main as the representation bewidely accepted political is,to areseen bycitizens That parties representation functions as itis supposed toin normative democratic theory (2007 p.169). party that theassumption on are based 1969) overtime (Converse identification voter ideathe thatpromote As arguedbyMainwaring those theories Zoco, and of strengthening for finding? this weaccount can then,How, significance. levels of statistical highest the reached variables factor. andIn both multivariate determining multi-level the models thecorresponding but rather that it is the period oftime in which democratisation took place that is the 54 CEU eTD Collection own representatives in key offices” (Katz & Mair 1995 p.6). Parties are seen as the only demands of the state, and ultimately attempt to capture control of the state by placing their partiesof aretraditionally defined whose social products goalhasmovements “make been to differentrole informing andstructuring society. In olderthe 1 democracies (groups and 2), vast majority partiesof in democracies post-1974 the andhave played a fundamentally in Parties tovoters haveafundamentally relationship democracies different older than the 2007 p.170)in keeping with periodthe effect. longevity. This finding wouldindicate rolethe of path dependence (Mainwaring andScully more than stablefor areason other are in oldersystems party democracies the conclude that Given the results of the multivariate regression and the table presented above, we must ** p<0.05;***p>0.01 etc. election third election, levels in election figuresand between volatility first of second correlation Spearman’s The row and column headings denote the number of the election for each country. Cells show Table.10 2 1st 3 6 5 4 nd rd th th th - Correlations of electoral volatility between between volatility elections of - Correlations electoral 37 0.410*** 51 0.4179*** 0.646*** 65 0.6239*** 0.723*** 68 1.000 0.706*** 68 0.806*** 68 N 0.359 1.000 1st 0.222 1.000 2nd 0.636*** 0.392** 0.606*** 0.302 3rd 55 .7 0.165 0.274 1.000 1.000 4th 1.000 5th 6th CEU eTD Collection (Mainwaring andZoco2007p.169). Consequently, elites have ashort time-horizon,moving in terms of support and finance, elites have little reason to remain faithful partyto labels (Mainwaring and Scully Onthesupply-side,1995 p.386). duetotheweakness parties,of both suggestion “partiesbeing only that of vote want the people,the nottheir opinion” exploitelectoral inorder topursue chancers to willing theirdemocracy ends, own the Furthermore, there is a distinct lack of trust felt towards parties with party elites often seen as (Katz and Mair 2009 p.755), or are made up of fragments of the former authoritarian elite. society, parties in new democracies often emerge through close integration with the state itself by electorate. Unlikethe parties in democraciesold thatrepresentspecific social in groups rolein heldin arenot avery and society same esteem have often different parties played the In countriesthose democratisedthat in later periods, after 1974,butparticularly 1990, after and institutionalised. stable well remains system party andthe arena, theelectoral enter parties for poorperformance and consequently volatility low,remains new parties donot their forgive will voters that mean bonds strong The electorate. the and parties between bonds membership and other party organs such as party media that have further strengthened these democracies therehas been a strong(although seemingly waning) emphasis on party allegiances were born that passed from generation to generations. Furthermore, in the older disenfranchised to gain the vote (Katz & Mair 1995 p.8). The result being that firm parties infact universal pre-date suffrage grew andoften out of strugglethe of the old democracies have with society. Oftentimes, particularly in the socialist/mass party model, One of mostthe important factors for influencing volatility is the strong relationship parties in established1998 p.599). (Toka parties entrepreneurs andinterest groups have no choiceother thantoseek their through goals argues, in the older democracies with high levels of party system institutionalisation, political captureviable means and state influencingmeansof bestthe for Toka policy; government as 56 CEU eTD Collection parties party whilst candidates areableuse modern to campaigning techniques andthe mass remain cynical of In these cases, voters 1997p.179). (Mair regime non-democratic previously defeata of or as collapse the a result of occurred in democratisation which democracies with voters that continue tobe passed on to this day. This is not the case in the post-74 parties were inherently tosocialtied helped groups, shape society and createdpowerful bonds explaining levels of party system institutionalisation. In the golden age of party development, Put simply, the period in which democratisation occurred in a country is a critical variable in exposure” (1995p.471). unknown candidates shotto top the of in pollsthe measure “thanks good totelevision argue, in presidentialthe elections in both Brazil in1989 and Peruin 1990, previously careers of political outsiders with virtually no party affiliation. As Mainwaring and Scully the in launching role key a playing television of examples well-known several are There dramatically. diminish elites the of behalf on fidelity party or building party for incentives the members (Sartori 1989 in 2007Mainwaring and p.167;Mair Zoco Consequently 1997p.176). messages widely needthe without for well-developed party due-payingorganisations or spread their areableto hand,candidates other massthe Innewon media. age of democracies the in before well society rooted became parties olderthe democracies, In democratisation. isnew mass democracies thepervasion of media, period the during particularly of television, in institutionalisation lowof system party tothe levels contributes that factor An additional stay faithful. to reason programmatic democratising countries, when voters judge parties for perceived failures they littlepossess degree “unprecedented of social Unlikedestructuring” (Mair 1997p.50). in earlythe inprogrammatically weak andheterogeneous terms of social makeup an havingexperienced levels of electoral volatility. Furthermore, many parties in the post-1974 democracies are between parties appearand whilstnewparties disappear–arecipefor parties established high 57 CEU eTD Collection matters for democracy for debate. for democracy remains open matters not what they 2001inone were”(Schimtter 2007 Mainwaring &Zocop. 166),whether this “parties fair are say seems that It to party organisations. depose essential once of the media to 58 CEU eTD Collection Hagopian, Gervasoni& Moraes 2008 p.364). (Mainwaring & Torcal 1995 p.391;Jones &Mainwaring 2003p.147;Epstein 2009 p.341; 5 that arguing scholars several has been seen as a textbook case of both party and party system underdevelopment, with This should come as little surprise for party system scholars given that for many years Brazil Figure.8 shows an obvious outlier, observation 196 -Brazil: we consider the multivariate models presented above, the Cook’s Distance score in model 1 in earlier presented selectthis paper tobetter fora divergent country further If examination. results utilise it statistical the is to possible on nested-analysis, literature the In keeping with institutionalisation, this paper will also consider amorein depth case study. As mentioned above, alongside a quantitative analysis of causes of party system 5 Case Study There were no such overtly identifiable outliers for model 2 formodel outliers overtlyidentifiable nosuch Therewere - Cooks distance- Cooks for model 1 ceteris paribus, ceteris itis among the weakest in Latin America 59 5 CEU eTD Collection access to power. Previous to this, parties failed to form any sort of programmatic agenda, and 1830s onwards, itwas not until party 1946 that competition became the p.355). Although there were formally political parties that competed for government from the undergone somany changes in radical party system makeup (Mainwaring andScully 1995 between the early 1800s and the presentday, no with other country in Latin America having systems party distinct seven been have there that argue example for Scully and Mainwaring 1998 p.172) has that had powerful consequences for sedimentationthe of party the system. The democratic history of Brazil is marked by a discontinuous pattern of evolution (Coppedge democratic rule, and more specifically, party rule in Brazil. to context historical some provide to is necessary it is firstly it question answer this to order In How then can we account for such a poorly institutionalised party system? p.356). andhas limitedseverely the quality democracyof in Brazil (Mainwaring andScully 1995 problems fordemocratic created hashelped governance, an inegalitarian social sustain order, 2008 p.365).As several scholars argue, this inordinately high level of fragilityparty has &Moraes Gervasoni (in Hagopian, party allegiances mentioned candidate’s the respondents candidate’s(33%), issue record positions (32.5%), and personal qualities (17%).Only 7% of National Election Study cited thefollowing considerations in voting for federal deputies: the consistently fail identifyto with parties. In 2002, for example, respondents to the Brazilian Furthermore, the poorly institutionalised party system can be observed by the fact that voters between chamber 1986 and 2006(making Brazil country in most fragmented the dataset). the Brazil is the level of fragmentation with an average 9.28 (effective) parties sitting in the lower centcompared21.92 to percent); but most the notable figure concerning theparty system in Electoral volatility in Brazilis markedly for theaverageabove group 3 (average 38.62per 60 major means of CEU eTD Collection government ARENA and the opposition by a “hegemonic two-party system” (Mainwaring and Scully 1995 p.355)made upof pro- the Following the military incoup 1964, thisincipient party system was anddissolved replaced election would atone and thenext emerge & Kinzo(D’Alvavanish 1999 p.143). then until andhowever, PTB), by large,and this period was characterised byfly-by-nightparties that democracy did begin toform, with three parties repeatedly contesting elections (PSD, UDN 1995 pp.356-358). During theperiod between 1945 –1964 an embryonic multiparty 1945 had already over adecade ofmassexperienced & (Mainwaringparty politics Scully beto contrasted Latin other American regimes such as Argentina,Chile andUruguay by who susceptible to anti-party the whims of presidents and political other leaders. This situation can circled elite,that the weresimply instead groups around constantly poorly organised simply obtaining patronage for their constituencies. The result of these two factors being that (Mainwaring andScully the 1995 p.368), main function of and parties politicians became legislative andlargely authority, arena executive electoral posts” the to restricting function whilst simultaneously,“reducing their [parties] legislative powers,expanding institutionalised party systems. Secondly,in allowing parties to even exist and Congress to 1979, the military regime crippled any development of party identities, aprerequisite of well- Firstly,in abolishing the old parties and allowing creation for of new parties in both 1965 and significantly undermined thelong term potential for party system institutionalisation in Brazil. Mainwaring notetwokey and Scully factors military(2005) of the have that regime in 1979, replacing them fivewith much smaller organisations. rising anti-regime sentiment, the military government again dissolved the two existing parties enough attracted support…thatit could gainsome legitimacy” (1995p.363).Afraid the of not involve the key decision-making positions within but military government, the the Mainwaring and Scully note, during this period “elections were not free, and contestation did Movimento DemocráticoMovimento Brasileiro 61 (MDB). As CEU eTD Collection directly for acandidate, or for aparty’s entire label. As Samuels argues,from a candidate’s – Open List Proportional Representation (OLPR). Under OLPR Brazilian voters can vote isit worth explaining theparticular system in employed elections totheChamber of Deputies individualism and weak allegiances party to organisations. Tounderstand why is this case,the p.488; Epstein 2009 p.335),whoargue itsthat mechanicsinherently politicalpromote particularly from party (Hagopianscholars al.et 2009p.363; Samuels 1999;Samuels 2000 (the lower house of the National Congress) which has come under severe criticism, for the Chamber of Deputies in system electoral place the in Brazil, specifically arrangement One of the most salient causes of such ineffectual parties is a result of the institutional bemust also several that considered. legacy, factors are to historical other there attributed Although there is a powerful argument that the inchoate party system in Brazil can be countries. Latin American with when compared other particularly era, in with contemporary parties the form lasting to bonds onthe ability citizens ruleof influence hashadapowerful democratic As we can see, the legacy of poorly institutionalised party systems in earlier stages of electoral options. their completely reassess requiredto were therefore other countries in the region, Brazilian citizens had no reference points on which to draw and country’s experience with political atparties point. Equally, this demandthe on side,in unlike bestthe means for engendering societal change –an incredibly difficultgiventask the that theelectorate andthey to convince in more provide merecould factwere than patronage labelsparty attempting and torecruitsuitable Additionally, candidates. were parties required democratised Brazil, therefore, party organisations were required to start afresh, creating new form or another, none of the pre-1964 parties existed in the New Republic. In the newly in one rule authoritarian survived had that countries American Latin in other unlike Brazil in 62 CEU eTD Collection garner votes andgarner votes strength the of their (Mainwaring parties and Scully 1995 p.359; D’Alva and specifically for their campaigns indicating a significant dissociation between their ability to presidential candidates have run on an open anti-party platform or have created parties presidency…encourages the p.145). Several personal (1993 campaigns” successful “thevery individualsAs D’AlvaandKinzocomment, are asked fact thatvoters for to choose institutional inarrangement Brazil is through negativethe impactof presidential competition. An additional way in which primacythe of party democracy is undermined by the connect seekcandidates non-party means to other with little surprise that potential voters. candidates $200,000 on campaigning costs and a federal deputy $600,000 (Mainwaring and Scully Paulo 1995p.381) Scully Paulo and (Mainwaring federal deputy$600,000 and a costs on campaigning $200,000 6 areamongcampaigns most the expensive in world the Given television that penetration in Brazil 95%inreached 2012) and2010 (Teleco political vehicle before theparties became well institutionalised (Mainwaring andScully 1995 p.396). electoral itself as animportant in established suchas Brazil television where particularly cases parties, political to challenge formidable a particularly pose can media mass above, mentioned by in strengthened roleis further the of mass mediaBrazil As in competition. electoral to voteforoneparty This (Epstein of nature another” over 2009 p.348). personalistic politics possible towin elections whichfurther undermining parties by “removing policy as a reason – tradingthe for of votes policy issues (Samuels as 1999p.419), candidates utilise any means is1999 p.419).There argument astrong that situation clientelismthis promotes andpatronage simply must rely on something other than the party label if they wish to get elected (Samuels candidates district, and per-party per multiple candidates are there thefactthat due to simply, always prefers a vote for himself or herself over a party label vote (Samuels 1999 p.495). Put therefore, candidate, plus votes…Each its party-label votes sum candidate’s of party’s the perspective, “the incentives for individualism are clear: The party’s total list vote equals the Mainwaring and Scully suggest that a successful state deputy in 1986 would need to spend around needtospend statedeputyin1986 would successful that a suggest andScully Mainwaring (Mainwaring andScully 1995 p.381;D’Alva &Kinzo 1999p.144),itcomes as 63 6 with almost no party funding for CEU eTD Collection their party’s programmatic agenda. Consequently,citizens arenolonger able toclearly to little reference with platform campaign on a personalistic areable to candidates presidential The role of the president in Brazil undermines party system institutionalisation in that firstly, Columns 3–Percentage of in seats Senatethe and Chamberof Deputies governorships of Columnstate wonby 2-Percentage each party Column 1 – Percentage of the election results in the first round presidential election D’Alva and Kinzo1993 p.145) Table.11 (Election2012). Deputies Resources Dilma Rousseff’s Partido dos Trabalhadores received on 16.9% of the seats in the Chamber of up until most2010 (the recentelections in Brazil)where the successful presidential candidate, in the Chamber of Deputies is not specific to the Collor de Mello period but continues to be It is important tonote, thisthat in situation which the President’sparty is poorly represented andfrom solicitsupport PRN opponents, creating mass-confusion for the party’s supporters. effectively, Collor deMello was therefore required to retreat from a number of party positions country the hadless than 10% safe andstable party supportin parliament. In governorder to Deputies (the lower house) the following year, resulting in a situation where the president of Chamber of the the PRNdidnot to for transfer in success below,can this table the see only in 22 representatives Congressthe of (D’AlvaDeputies andKinzo1993 p.146). As we officially represented the Partido da Reconstrucao Nacional (PRN), a young party that had ideology they were not natural bedfellows, the eventual winner, Fernando Collor de Mello phenomenon and the consequences it can cause for the party system. Although in terms of Kinzo1993 p.146).The 1999presidential election provides a clear example of this Parties PRN Position of the Parties in Executive and Legislative Office (%) (adapted from (adapted (%) Office Legislative and Position in Parties of the Executive President1989 28.5 1 Governor 1990 0.0 2 64 Senate 1990 Senate 3.7 3 Fed. Chamber 1990 8.0 3 CEU eTD Collection (adapted from (adapted Epstein 2009 p.340) Figure.9 to Chamberthe between of 1982andDeputies 2002. & Paiva The 2006p.5). tablebelow party showsthe shares for theRiode Janierodelegation levelsin of government), 1990 numberthis jumped had tothirty-three (Nicolauin 1998 Bohn Paiva note that in 1986, twenty-nine parties took part in the national elections (across all high, comparatively speaking, astoalmost putBrazil off map” the (2000 p.241).Bohn and world and certainly the highest of any countries in the dataset – as Samuels puts it, “almost so in the severe most the among is Brazil in fragmentation of level the above, mentioned As parties in tosuccessfullyorder (D’Alva andKinzo1993 p.145).govern political parties due to their need to interact and win the support of a wide range of other from their autonomy have presidents considerable legislative seats, to success executive identity. Secondly, due to the fact that the president’s party rarely,if ever, transfers their distinguish whatit is that a particular party represents and therefore form any sort of party - Party - Party shares Riode Janeirodelegation to Chamber of Deputies 1982–2002 65 CEU eTD Collection historical legacy historical repeated destruction and the authoritarianism of of any party developing extremely poor levels of party system institutionalisation in Brazil. Undoubtedly, the tothe have contributed factors that awideof are variety we canseethatthere In conclusion, post. lasting bonds, with their representatives who are liable to leave office at any point for another This results in a situation in which citizens are left unable to, or at least reluctant toform placesincumbents names onthe subsequent ballotautomatically (Samuels 2000pp.482-484). municipal governments, despite the fact that a ‘birth right candidate law’ ( seeking25.6% positions outside average, only 74.4% of incumbents have run for re-election in Brazil with the remaining incumbents in the Chamber of Deputies choose not to contest the subsequent election. On in is, of Brazil; that number ahuge turnover legislative problem with an enormous fact that citizens form relationships with personalities as opposed to parties. Similarly, there is faithful to aparty for electoral success, and also acause of the very same weakness due to the p.419). This is both a sign of the weakness of parties – elites do not feel they have to remain and eventoideological non-contiguous parties (Hagopian al.et 2008p.364; Samuels 2000 1998 legislature, 137 of 513incumbentsthe (27%)crossed the floor,many more than once 1994 legislature for instance, 207of 504incumbentsthe (41%)changed parties. Inthe1995- As Samuels levelthe reports, floor-crossingisof almost in unprecedented Brazil. Inthe1991- of party fidelity part representatives. the lackon party and complete the manoeuvrings elite in is system party Brazil the to thehasfragility of contributed factor that Finally, aparticular sort of relationships with these parties and also to hold them to account for their actions. election on election, others come and go making it impossible both for citizens to form any –an period astonishing figure. Although some parties consistently some receive votes this during chamber lower tothe elected were parties of seventeen delegates As wecansee, the Chamber such as in the much more powerful state or 66 candidato nato ) CEU eTD Collection emerging inemerging nearthe future. they are trying to woo, there seems to be little hope for awell institutionalised party system political parties on a day-to-day basis cannoteven commit tothem never mind the electorate Brazilian case and alsomost worrying for the future – if those who are meant to represent one party label and another at an alarming rate. It is this fact that I find most striking in the seem valueto party government, but party elites themselves who are willing to jump between successfully. Finally,it is important to note that itis not only voters in Brazil who do not who is essentially forced to engage in ‘wheeler dealer’ or scheming tactics in order to govern driven politics (often channelled through the mass media), and the role the role of a president personality- on emphasis the particularly in Brazil, politics electoral of structure mechanical in other Latin American countries. Similarly, stabilisation has been undermined by the system meant that citizens were unable toform the long-term relationships with parties seen 67 CEU eTD Collection more likely to switch between parties who fail bedto any kind of significant roots in society. therefore and are party programmes between differentiate unable are to voters fragmentation, determining levels of party system institutionalisation. In party systems with a high degree of Unsurprisingly,levels fragmentation of were consistently found play to a key in role all across (2007) fourMainwaring Zoco and wavesof democracy. more recently,isinaugurated particularly notable and as it theconclusionsextends of regimes those level lower than volatility of electoral a experience significantly periods earlier in democracies that democratisation within thatinaugurated analysis. Thissuggests the those ‘period effect’ thesis (Mainwaring& Zoco 2007p.166) by including afourth wave of ratherbut periodthe in which democratisation took place. Thisfinding substantiates the institutionalisation, system party of levels to contributes that contested elections of number the or ageof a democracy is not itthe that abovesuggest presented it, results the before hasgone that literature tomuch of the here.contrary Similarly, out other each cancelling effects two are chance there is that distinct that a given there research for future fruitful area likely none atall. Of twovariables,the arguablylevels of education would provide themore party system institutionalisation, Ifound that they have at best negligible effects, and more had both level suggested that of hadeducation and adetermining GDP levels on effect of literature, thereby opening upavenues for further exploration.Whilst upuntil now, scholars Interestingly, some ofmy findings what contradict has beenpreviously concluded by the analytically rigorous. considered before, findings the above presented would appeartobe both robustand been ever has than cases of range a wider to analysis the extending In institutionalisation. party factorswhich to contribute of those comparative understanding system In conclusion, this study has achieved its broad aim of expanding the frontiers of the 6 Conclusions 68 CEU eTD Collection institutionalisation should considermore prominently suchsupply-side factors as rolethe toensureorder thorough analysis,future studies intocauses of party system theories that have gone before it, but also open up new opportunities for further research. In In brief, the findings of this study not only go some way to substantiate (or weaken) those party elites have by largeand not formed relationships strong with either.parties electoral volatility and enormous levels fragmentation; of but arguably more importantly, citizens have failed toform strong relationships parties,with resultingin levelshigh of side, that is, the actions of voters; we must also consider the actions of party elites. In Brazil, those factors that influence party system institutionalisation are not restricted to the demand Additionally, Iwould argue that the key aspect to emerge from the study of Brazil is that they have difficulty intotapping the potentially effects decisive of historical legacy. Brazil case shows asthe institutionalisation, party can system to contribute that factors the qualitative analysis.Although the statistical models seem provideto a robust explanation of The case study of Brazil provided a deeper level of analysis and emphasised the benefits of hierarchical control (Protysk 2006; Cheibub & Chernykh 2009). 2006 p.219)and resulting the confusion between accountability, and responsibility of patterns (Protsyk coexistence with ‘intra-executive’ associated complications inevitable seemingly in the lies countries in these in volatility increase this of account plausible most the tentative, significant across all three models. Although my explanation of this phenomenon remains substantively statistically and both itremained fact the notable that dueto particularly is This volatility. electoral and systems semi-presidential between link direct seemingly the paper tonote firstis this far the am as aware, asI results, comparable somewhat volatility.Although there have been a handful of studies on similar topics that have found presidential regimes, semi-presidential regimes directly contribute to higher levels of electoral Finally, arguably mostthe unique finding of analysisthe is thatunlike parliamentary or 69 CEU eTD Collection particularly fascinating but enigmaticbut fascinatinginparticularly study element the of democracy. contemporary continue provide to fruitful and results helpfuture scholars pin down what remains a field tocontribute has attempted to the can byopeningupnewavenuesfor research that To conclude, party systems remain an important area of democratic scholarship. This study institutionalisation in a wider range of electoral contests such as those at a local or state level. natural extension of this thesis would be toconsider the causes of party system presidentialism on electoral volatility would surely warrant more in depth study. Finally, a strength of the relationship shown in the statistical analyses, the influence of semi- elites can play in undermining strengthening or party system Furthermore,stability. given the 70 CEU eTD Collection Casal F.,& Bértoa, Enyedi, “Patterns Z.(2011). of Inter-Party Competition (1990-2009)”in Casal Bértoa, F., & Enyedi, Z. (2010). Party System Closure – Conceptualisation, Casal Bértoa, F. (2010). Sources of Party System Institutionalisation in New Democracies: Carothers, T. (2006). Confronting the Weakest Link : Aiding Political Parties in New Burnell, P.,& Gerrits, A. (2010). Promoting Party Politics in Emerging Democracies. Bohn, S. R., & Paiva, D. (2004). The socio-economic basis of the electoral volatility of recent Birnir,PublicJ.K. (2005). Capital Venture and Party Institutionalization. Comparative Birch,(2003). Electoral S. Systems and Political inTransformation Post-Communist Europe. Basedau, M.,&Stroh,A. (2008). Measuring Party inInstitutionalisation Developing Bartolini, S.,& Mair P.(2001). Challenges to Contemporary Political Parties. In Political Bader, M. (2010). Party politics in Georgia and Ukraine and the failure of Western Allbright, J. (2009). Does political knowledge erode party attachments?: A review of the Ágh, A.(1998).The end of beginning:the partial the consolidation Centralof East European References Cienca Politica. Cienca de 3 CongressoLatino-Americano from caseof The Brazil. Proceedings democracies. Palgrave MacMillan. Basingstoke: cognitive mobilization thesis. Electoral 28,248 Studies, - 260. Paul G. Lewis and Rados https://disc.ceu.hu/sites/default/files/field_attachment/page/node- Operationalisation and Validation. Working Disc Series. Paper Accessed 31/3/2012. http://cadmus.eui.eu/handle/1814/15804 Lessons from Central East EUISPSWorkingEurope. Papers.Accessed 31/3/2012. Democracies. Washington D.C.: Carnegie Endowment for International Peace. 1065-1084. Democratization, 17(6), Political 38(8),915-938. Studies, Institute of Global Area Working Studies, Paper. Countries: A New Research Instrument Applied to 28 African Political Parties. German and London:The JohnsHopkins University Press. Parties and Democracy, eds.Larry Diamond,Richard and Gunther, Baltimore 328-342. Assistance. Democratization, 1085-1107. 17(6), Lane, London: 202-216. Routledge. Party Systems. In University Press. University Manchester Manchester: Europe, andCentral on Eastern Perspectives Comparative 3320/discwp201011.pdf Comparing Party System Change, eds. Paul Pennings, and Jan-Erik á aw Markowski (eds.) Europeanizing Party Politics? 71 CEU eTD Collection Election Resources. (2010). Election Resources on the Internet: Federal inBrazil. Elections on Election Internet: the (2010). Resources Resources. Election Duverger, M. (1954). Political Parties: Their Organisation and Activity in the Modern State. Gallagher, M. Election Indices. (2012). Available at Dolidze, V. (2005). Political Parties and Party Development in Georgia. Central Dassonneville, R., & Hooghe, M. (2011). Mapping Electoral Volatility in Europe. Dalton, R. J., McAllister, I., & Wattenburg, M. (2000). The Consequences of Partisan Dalton, R.J. (1984).Cognitive Mobilization and Partisan Dealignmentin Advanced Coppedge, M. (1998). The Evolution of Latin American Party Systems. In S. Mainwaring & Converse, P. (1969). Of Time and Partisan Stability. Comparative Political Studies, 2,139 - Cheibub, J. A., & Chernykh, S. (2009). Are semi-presidential constitutions bad for democratic Casal Bértoa, F., & Mair, P. (Forthcoming). Party system institutionalisation across time in Hicken, A., & Kuhonta, E. (2011). Shadows From the Past: Party System Institutionalization Hale, H. (2006). Why Not Parties in Russia? Democracy, Federalism and the State. Hagopian, F., Gervasoni, C., & Moraes, J. A. (2008). From Patronage to Program: The Ferree, K. E. (2010). The Social Origins of Electoral Volatility in Africa. British Journal of in Brazil. Development Party of Problems Ideology: Versus Clientelism (2009). J. D. Epstein, New York: Wiley. York: New Bulgaria. Sofia, Research, Electoral ConferenceComparative on European from 1st Proceedings Oxford: Oxford University Press. Dealignment. In R. J. Dalton & M. Watternburg (Eds.), Parties without Partisans. Industrial Journal Democracies. of 46(2), Politics, 264 -284. A. Valenzuela (Eds.), Politics, Society, and Democracy. Boulder: Westview Press. 171. performance? Constitutional 202-229.Political Economy, 20(3), post-communist Party Europe. Government, 85-112. Cambridge: Cambridge University University Cambridge:Press. Cambridge 42(3), 360-391. Political Studies, Comparative in Brazil. Legislators of Party-Oriented Emergence Political 40(4),759- Science, 779. Available at Available Caucasus, 249–59. in Asia. Comparative Political Studies Political Comparative in Asia. ndices.pdf http://www.tcd.ie/Political_Science/staff/michael_gallagher/ElSystems/Docts/ElectionI Party Party 335 Politics, 15(3), - 355. , Accessed May 2012. http://electionresources.org/br/index_en.html , 44(5),572-597. 72 , AccessedMay 2012. Asia and the and Asia CEU eTD Collection Lijphart, A. (1994). Electoral Systems and Party Systems: A Study of Twenty-Seven S.Lieberman, E. (2005).Nested Analysis aMixed-Methodas Strategy for Comparative Levitsky, S.(1998). Institutionalization and Peronism. the Concept, the Case and the Case for Lewis, P.(2005). Party systems in Post-Communist Central Europe: Patterns of Stability and PoliticalR. (2007). AHeterogeneous Electorate: Predisposition Lachat, Sophistication, Laakso, M., & Taagepera, R. (1979). Effective Number of Parties: A Measure with Kubicek,(2009). Problems P. of theOrangePost-post-communism: Ukraine after Revolution. Kitschelt, H., & Smyth, R. (2002). Programmatic Party Cohesion in Emerging Post- Kitschelt, H. (2001). Partisan Competition and Welfare State Retrenchment: When do Kinzo, M.D. A. G. (1993). Brazil: The Challenge of the 1990s. London: The Institute of Katz, R. S., & Mair, P. (2009). The Cartel Party Thesis: A Restatement. Perspectives on Katz, R. S., & Mair, P. (1995). Changing Models of Party Organisation and Party Democracy: Jones, M. P., & Mainwaring, S. (2003). The Nationalisation of Parties and Party Systems: An Jones, M. P., & Mainwaring, S. (2003). The Nationalization of Parties and Party Systems: An Jackman, R.W.(1987). Political Institutions andTurnourVoter in Industrialthe R. Inglehart, (1970). Cognitive Mobilization Identity.and European Comparative Politics, Huntington, S. P. (1968). Political Orderin Changing Societies. New Haven, CT: Yale Consolidation. Democratization, 562 13(4), - 583. Nomos. Baden-Baden: and Process. Voting the Strength State (pp. 265 - 305).Oxford: Oxford University Press. Politicians Choose Unpopular Policies? In P. Pierson (Ed.), New Politics of the Welfare Studies. American Latin 753 Politics, 7(4), - 766. 5-28. 1(1), Politics, Party Cartel Party. of the The Emergence Empirical MeasureApplication and Americas.to the 9(2),139 - 166. Empirical Measure and anApplication to the Americas. Party Politics, 9(2), 139 - 166. Democracies. American Political ScienceAssociation, 405 81(2), - 424. 3(1), 45-70. Research. TheAmerican Political Review,Science 99(3), 435-452. Unpacking the Concept. Party Politics, 4(1), 77-92. ApplicationWest to Europe. Comparative Political Studies, 12(1), 3 -26. Democratization, 323– 16(2), 43. Studies, 35,1228-1256. Political Comparative Context. RussiainComparative Democracies: communist Press University 73 CEU eTD Collection Randall, Vicky, and Lars Svasand. (2002). Party Instituionalization in New Democracies. Protsyk,(2006). Intra-Executive O. Competition between Presidentand Prime Minister: Powell, E.N. & Tucker, J. A. (2009). New Approaches to Electoral Volatility: Evidence from Pennings, P.,& Lane, J. (1998). Comparing Party System Change. In Comparing Party Pedersen, M. N.(1979). The Dynamics of European Party Systems. European Journal of Netherlands Institute for Multiparty Democracy (2004).A Framework for (NIMD). Citizens Leaders, Kyrgyzstan: in Post-Soviet Democracy Transition to (2011). D. Murzaeva, Mainwaring, S.,& Zoco, E. (2007). Political Sequences and the Stabilisation of Interparty Mainwaring, S., & Torcal, M. 2006. ‘Party System Institutionalization and Party System Mainwaring, S.,& Shugart, M. S. (1997). Juan Linz, Presidentialism and Democracy: A Mainwaring, S., & Scully, T.R. (1995). Introduction: Party Systems in Latin America. In Mair, P., & Casal Bértoa, F. (2010). Two Decades On: How Institutionalized are the Post- Mair, P. (1997).Party System ApproachesChange: and Interpretations. Oxford: Clarendon Mair, P.(1993). Myths of electoral change and the survival of the traditional parties. Studies, 54, 219 Studies,- 244. 54, Patterns of Institutional Conflict and Cooperation under Semi-Presidentialism. Political Political 1-26. Research, 7(1), Critical Appraisal. Politics, Comparative 449 29(4), -471. Press. European Journal of Political 121 Research, 24(2), - 133. Democracies 1945-1990. Oxford: Oxford University Press. System Change, eds. Paul Pennings, and Jan-Erik Lane, London: Routledge. Democracy. Multiparty Institute for Netherlands Handbook. Development Institutional Building. Party Democratic Publishing. Academic LAPLAMBERT Saarbrücken: Legitimacy. of Political and Perceptions 155 -178. Competition: Electoral Volatility in Old and New Democracies. Party Politics, 13 (2), Katz Richard S. and William 204–27. Crotty, pp. London: Sage. Theory After the Third Wave of Democratization’,In Handbook of Political Parties, eds. Press. Mainwaring, Timothy and R. Scully,California: Stanford, University 1-37. Stanford Building Democratic Institutions. Party Systems in Latin America, eds. Scott CommunistParty EUIWorkingSystems? Papers,1-26. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1449112 SSRN: Postcommunist Countries. APSA Meeting2009 Toronto Paper.Availableat http://ssrn.com/abstract=1449112; 74 CEU eTD Collection Wibbels, E.,& Roberts, K. M. (1999). Party Systems and Electoral Voting in Latin America: Unesco Institute forUnesco Institute StatisticsGlobal (2010). Digest 2010:ComparingEducation Education Tóka, G. (1998). Party Appeals and Voter Loyalty in New Democracies. Political Studies, Teleco, “Rádio eTV”, (2012). Report May 2012. Available at Tavits, M.(2008).On linkagethe between electoral volatility and party instability system in Tavits, M. (2005). The Development of Stable Party Support: Electoral Dynamics in Post- Exploration Adults: among Identification Party of Development The (1979). P. W. Shiveley, Schimitter, P. C. (2001). Parties Are Not What They Once Were. In Political Parties and Sartori, G.(1990). A typology of party systems. InThe West European Party System, ed. Samuels, D.J., & Shugart, M. S. (2010). Presidents, Parties, and Prime Ministers: How the Samuels, D.J. (2000). The Gubernational Coattails Effect: Federalism and Congressional Samuels, D. J. (2000). Ambition and Competition: Explaining Legislative Turnover in Brazil. Samuels, D.J. (1999). Incentives to Cultivate a Party Vote in Candidate-centric Electoral Rose, R., & Mishler, W. (2010). A Supply-Demand Model of Party System Science 575 Review, 93(3), - 590. A Test of Economic, Institutional, and Structural Explanations. The American Political 46(3), 589-610. Central and Eastern Europe. European Journal of Research, 537 Political 47(5), - 555. Communist American Europe. Journal Political283 of Science, 49(2), - 298. of a Functional Model. American Political Review, 1039- Science 73(4), 1054. Cambridge University Press. Separation Affectsof Powers Party Organisation andCambridge: Behaviour. Elections in Brazil. TheJournal Politics,of 62(1), 240 -253. Legislative Quarterly, Studies 481 - 497. 25(3), Systems: Evidence from Brazil. Political 487 Comparative Studies, 32(4), - 518. Institutionalization: The Russian Case. Party Politics, 16 (6), 801 - 821. http://www.unesco.org/education/docs/EN_GD2004_v2.pdf Statistics Across the World. Available at May2012. Accessed http://electionresources.org/br/index_en.html, The Johns Hopkins University Press. Democracy, eds. Larry Diamond, and Richard Gunther, 67-98. Baltimore and London: Mair,Peter 316-349. Oxford: Oxford University Press. Party Politics, 8(1), 5-29. 75 , Accessed May 2012 CEU eTD Collection Countries includedin and volatility dataset scores Appendix 1 Czech Republic 2 Czech Republic Montenegro2 Great Britain Great Luxembourg Herzegovina Macedonia Dominican Costa Rica Costa Slovakia 2 Slovakia Argentina Colombia Botswana Lithuania Denmark Australia Republic Romania Moldova Hungary Armenia Slovenia Belgium Bulgaria Ecuador Country Serbia 2 Ukraine Georgia Albania Bosnia- Estonia Canada Bolivia Austria Iceland Croatia Poland Russia Latvia Brazil Malta Chile Period consideredPeriod 1994 -2009.7 1978 -2006 1946 -2004 1994 -2007 1993 -2007 1992 -2008 1979 -2002 1945 -2005 1986 -2006 1946 -2003 1992 -2008 1987 -2008 1990 -2008 1992 -2007 1990 -2006 1990 -2009 1991 -2009 1989 -2005 1945 -2006 1983 -2003 1990 -2006 1990 -2007 1991 -2007 1989 -2004 1992 -2008 1993 -2006 1987 -2005 1990 -2006 1991 -2007 1946 -2006 1958 -2006 1966 -2004 1985 -2005 1946 -2002 1990 -2009 1987 -2009 1990 -2007 1990 -2008 1998 -2006 Average volatility 76 33.20714286 6.865217391 45.25333333 31.92888889 10.96086957 38.86333333 48.30333333 11.62368421 9.713333333 29.85533333 6.635294118 17.28833333 34.21333333 15.930625 10.62875 47.2725 45.5625 13.8775 69.2675 50.4675 10.3275 30.9975 45.718 36.405 19.622 11.665 56.644 46.925 26.728 22.484 42.115 46.034 26.425 39.452 30.198 49.296 2.346 58.35 55.61 (%) No. of elections considered 18 23 10 23 19 15 16 17 5 7 4 4 3 9 5 4 5 5 4 5 6 6 4 5 4 5 3 4 4 4 4 3 8 5 4 6 5 5 3 CEU eTD Collection Papua NewGuinea Papua New Zealand Trinidad and Netherlands Switzerland El Salvador Venezuela Sri Lanka Mauritius Honduras Germany Malaysia Uruguay Portugal Jamaica Norway Sweden Mexico Finland Tobago Turkey Ireland Greece France Spain Japan Israel India USA Italy 1974 -2004 1959 -2002 1949 -2003 1958 -2005 1945 -2005 1946 -2004 1983 -2002 1977 -2004 1994 -2006 1976 -1995 1952 -2005 1948 -2001 1948 -2002 1951 -2004 1981 -2005 1974 -2004 1949 -2005 1946 -2002 1945 -2003 1985 -2006 1947 -2003 1948 -2002 1946 -2003 1977 -1997 1984 -2004 1966 -2002 1952 -2004 1976 -2005 1946 -2005 77 13.27142857 11.27133333 3.419310345 14.11578947 15.35846154 26.74307692 6.788333333 8.043333333 18.72857143 17.84142857 7.439285714 7.929411765 27.34444444 16.72727273 16.14454545 17.59375 20.5625 8.03125 14.695 20.112 32.936 32.694 9.4625 10.833 15.595 11.087 19.26 12.51 27.75 10 15 10 15 16 10 29 19 13 13 15 14 16 14 17 20 17 11 11 7 5 8 4 5 6 7 4 4 9